This ChangeLog file is no longer maintained - see the git repo history for more recent changes: https://xapian.org/bleeding Wed Sep 30 19:34:51 GMT 2015 Olly Betts <olly@survex.com> * languages/compiler/: Merge changes from upstream snowball again. Sun Sep 27 23:59:05 GMT 2015 Olly Betts <olly@survex.com> * languages/compiler/: Merge changes from upstream snowball. Sat Sep 26 04:07:35 GMT 2015 Olly Betts <olly@survex.com> * tests/api_backend.cc: Remove pointless call to get_database(). Sat Sep 26 03:21:45 GMT 2015 Olly Betts <olly@survex.com> * common/pack.h,configure.ac,include/xapian/types.h, include/xapian/version_h.cc,tests/api_backend.cc,tests/api_wrdb.cc: Add new configure options --enable-64bit-docid and --enable-64bit-termcount, which control the size of these types. Currently the backends only support 32-bit docids, but using 64-bit docids allows searching multiple database where the mapped docids don't fit in a 32-bit type. Based heavily on patch from James Aylett and Dylan Griffith. Fixes #385. Sat Sep 26 00:32:31 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_table.cc: Use BLK_UNUSED constant. Sat Sep 26 00:31:45 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc: Initialise last_readahead - was causing testsuite failures under valgrind. Fri Sep 25 06:28:02 GMT 2015 Olly Betts <olly@survex.com> * net/length.cc,net/length.h: Provide overloads for unsigned, unsigned long, and unsigned long long, rather than uint32_t and uint64_t - this avoids overload resolution failing when we pass a type which isn't one which uint32_t or uint64_t is a typedef to. (see #385) Fri Sep 25 04:10:00 GMT 2015 Olly Betts <olly@survex.com> * api/vectortermlist.cc: Check decoded length. Sun Sep 13 08:27:53 GMT 2015 Olly Betts <olly@survex.com> * common/socket_utils.cc: Change C-style cast to reinterpret_cast. Wed Sep 02 03:13:45 GMT 2015 Olly Betts <olly@survex.com> * docs/stemming.rst,languages/armenian.sbl,languages/basque.sbl, languages/catalan.sbl: Update snowball website links to snowballstem.org. Correct "2-clause BSD" to "3-clause BSD". Sun Aug 30 04:43:49 GMT 2015 Olly Betts <olly@survex.com> * net/remote_protocol.rst: Markup tweak. Thu Aug 13 08:56:25 GMT 2015 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc: Improve error when the client and server remote protocol versions aren't compatible. Tue Aug 11 09:32:46 GMT 2015 Olly Betts <olly@survex.com> * docs/: Strip out "quickstart" examples which are out of date and rather redundant with the "simple" examples. Tue Aug 11 08:24:48 GMT 2015 Olly Betts <olly@survex.com> * pkgconfig/xapian-core.pc.in: Fix library name in .pc file to say "xapian" not "xapian-core". Reported by Eric Lindblad to the xapian-devel list. Tue Jul 07 01:37:00 GMT 2015 Olly Betts <olly@survex.com> * api/omenquire.cc,api/query.cc,backends/chert/,backends/database.cc, backends/database.h,backends/glass/,common/io_utils.cc, common/io_utils.h,configure.ac,include/xapian/enquire.h, include/xapian/query.h,matcher/multimatch.cc: Where posix_fadvise() is available, use it to prefetch postlist Btree blocks from the level below the root block which will be needed for postlists of terms in the query, and similarly for the record/docdata table when MSet::fetch() is called. Based on patch by Will Greenberg in #671. Mon Jul 06 09:44:54 GMT 2015 Olly Betts <olly@survex.com> * common/safeunistd.h,configure.ac: Don't provide our own implementation of sleep() under __WIN32__ if there's already one - mingw provides one, and in some situations it seems to clash with ours. Reported to xapian-discuss by John Alveris. Mon Jun 29 12:10:22 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_compact.cc: Need <unistd.h> for unlink() on FreeBSD, reported by Germán M. Bravo via github PR 72. Wed Jun 24 10:07:05 GMT 2015 Olly Betts <olly@survex.com> * tests/harness/testutils.cc,tests/harness/testutils.h: Use 'Xapian::docid' instead of 'unsigned int' so that things work with 64 bit docids (see #385). Tue Jun 23 13:38:54 GMT 2015 Olly Betts <olly@survex.com> * api/snipper.cc,api/snipperinternal.h: Use appropriate typedefs rather than unsigned int. Split out of patch by Dylan Griffith in #385. Tue Jun 23 13:32:36 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_types.h, backends/glass/glass_database.cc,backends/glass/glass_defs.h: Add constants CHERT_MAX_DOCID and GLASS_MAX_DOCID for the largest docid which the backends support, independent of the size of type Xapian::docid. (helps #385) Tue Jun 23 13:19:06 GMT 2015 Olly Betts <olly@survex.com> * tests/api_backend.cc: Add regression test for #678. Tue Jun 23 13:04:06 GMT 2015 Olly Betts <olly@survex.com> * api/,backends/remote/net_postlist.cc, backends/remote/remote-database.cc, geospatial/latlong_posting_source.cc,net/,tests/unittest.cc: Split decode_length() into versions which decode a 32 vs 64 bit value, and which check that there's at least the decoded length in bytes in the source data. (Fixes #678 and helps #385) Tue Jun 23 11:49:15 GMT 2015 Olly Betts <olly@survex.com> * net/remoteserver.cc: Check length of key in MSG_SETMETADATA. Sat Jun 20 13:05:50 GMT 2015 Olly Betts <olly@survex.com> * common/str.h,tests/,tests/harness/,tests/internaltest.cc, tests/perftest/perftest.cc,tests/perftest/perftest_matchdecider.cc, tests/perftest/perftest_randomidx.cc,tests/queryparsertest.cc, tests/soaktest/soaktest_queries.cc,tests/termgentest.cc, tests/unittest.cc: Make str() private to the library, and use C++11 std::to_string() to convert numbers to std::string in the testsuite. (ticket #63) Sat Jun 20 12:32:13 GMT 2015 Olly Betts <olly@survex.com> * tests/api_unicode.cc: Fix trailing whitespace introduced by previous commit. Sat Jun 20 06:31:27 GMT 2015 Olly Betts <olly@survex.com> * tests/api_unicode.cc,unicode/Makefile.mk, unicode/UnicodeData-README.txt,unicode/UnicodeData.txt: Update to Unicode 8.0.0. (fixes #680) Thu Jun 11 11:57:50 GMT 2015 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Remove XAPIAN_CONST_FUNCTION marker from sortable_serialise_() helper, as it writes to the passed in buffer, so it isn't const or pure. Fixes decvalwtsource2 testcase failure when compiled with clang. Thu Jun 11 05:05:27 GMT 2015 Olly Betts <olly@survex.com> * examples/xapian-metadata.cc: Extend "list" subcommand to take optional key prefix. Sat Jun 06 11:02:50 GMT 2015 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Use C++11 auto and container iteration for cleaner code. Fri Jun 05 12:08:02 GMT 2015 Olly Betts <olly@survex.com> * api/omenquire.cc,api/omenquireinternal.h, backends/remote/remote-database.cc,backends/remote/remote-database.h, include/xapian/matchspy.h,matcher/,net/remoteserver.cc, tests/api_none.cc: Add support for optional reference counting of MatchSpy objects. Fri Jun 05 02:55:54 GMT 2015 Olly Betts <olly@survex.com> * backends/remote/remote-database.h: Correct doxygen comment for internal method. Thu Jun 04 12:44:00 GMT 2015 Olly Betts <olly@survex.com> * examples/quest.cc: Prefer pre-increment ++. Thu Jun 04 12:42:49 GMT 2015 Olly Betts <olly@survex.com> * tests/api_compact.cc,tests/api_posdb.cc: Remove unused variables. Thu Jun 04 12:42:04 GMT 2015 Olly Betts <olly@survex.com> * tests/api_backend.cc: Fix compilation in case where S_ISSOCK() isn't defined. Thu Jun 04 12:41:31 GMT 2015 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Remove redundant assignment. Thu Jun 04 12:40:52 GMT 2015 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Remove extra "break;". Thu Jun 04 12:39:51 GMT 2015 Olly Betts <olly@survex.com> * include/xapian/query.h: Remove bogus extra "Query::" qualifier from fake Query constructor prototypes provided for SWIG. Thu Jun 04 12:38:36 GMT 2015 Olly Betts <olly@survex.com> * api/documentvaluelist.h,api/leafpostlist.h,api/matchspy.cc, include/xapian/,net/length.cc: Mark constructors taking a single argument as "explicit" to avoid unwanted implicit conversions. Mon Jun 01 13:11:02 GMT 2015 Olly Betts <olly@survex.com> * NEWS.SKELETON: Remove section for "flint backend" from template. Mon Jun 01 13:09:55 GMT 2015 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update for 1.3.3. Mon May 25 04:33:15 GMT 2015 Olly Betts <olly@survex.com> * NEWS: Update. Sun May 24 12:52:47 GMT 2015 Olly Betts <olly@survex.com> * tests/api_postingsource.cc: timer_create() seems to always fail on AIX with EAGAIN, so just skip the matchtimelimit1 testcase there. Sun May 24 11:48:30 GMT 2015 Olly Betts <olly@survex.com> * net/tcpserver.cc: Under __WIN32__, inet_ntop()'s second parameter is 'void*' (to POSIX says it should be 'const void*'). Sun May 24 11:46:25 GMT 2015 Olly Betts <olly@survex.com> * common/safenetdb.h: Under __WIN32__, MSDN says that newer platforms need <ws2def.h> for AI_* constants. And mingw doesn't seem to define AI_NUMERICSERV yet. Sun May 24 10:41:39 GMT 2015 Olly Betts <olly@survex.com> * common/safewindows.h,configure.ac: Under __WIN32__, we need to specify Vista as the minimum supported version to get the AI_ADDRCONFIG flag. Older versions seem to all be out of support anyway. Sun May 24 06:46:15 GMT 2015 Olly Betts <olly@survex.com> * net/tcpclient.cc: Add missing declaration of variable under __WIN32__. Sun May 24 06:39:56 GMT 2015 Olly Betts <olly@survex.com> * tests/harness/testsuite.h: Simplify FAIL_TEST and SKIP_TEST slightly. Sat May 23 11:52:03 GMT 2015 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc: Don't leave an extra fd open when starting xapian-tcpsrv for remotetcp tests. Sat May 23 10:12:31 GMT 2015 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc: Add spaces between literal strings and macros which expand to literal strings for C++11 compatibility in __WIN32__-specific code. Fri May 22 03:22:03 GMT 2015 Olly Betts <olly@survex.com> * NEWS: Update. Wed May 20 14:24:03 GMT 2015 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: If old and new __GXX_ABI_VERSION are both >= 1002 (which means GCC >= 3.4), then issue a warning about mismatching versions instead of an error. The changes in these versions are bug fixes for corner cases, so there's a good chance of things working - e.g. building xapian-bindings with GCC 5.1 (which defaults to __GXX_ABI_VERSION 1008) against xapian-core built with GCC 4.9 (1002) seems to work OK. A warning is still useful as a clue to what is going on if linking fails due to a missing symbol. Wed May 20 13:47:50 GMT 2015 Olly Betts <olly@survex.com> * net/serialise.cc: Don't use encode_length() on a bool - it's overkill, and leads to a warning from GCC 5.1 (due to the comparison (bool_value < 255) from the template expansion always being true). Wed May 20 01:49:18 GMT 2015 Olly Betts <olly@survex.com> * api/matchspy.cc: Fix comparison function not to return true for two equal elements, which fixes matchspy4 test failure with clang and libc++. Wed May 20 00:20:56 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: Disable "<FUNCTION> is expected to return a value" warning from Sun's C++ compiler, as it fires even for functions we end in a "throw" statement. Genuine instances will be caught by compilers with superior warning machinery. Mon May 18 04:18:37 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: Arrange to pass command line option so that xlC actually fails to compile the test code for typeid() when RTTI isn't enabled. Sun May 17 11:35:36 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_version.cc: When reporting an error that the glass format version doesn't match, don't append an uninitialised char[] buffer to the error string (code accidentally left behind by the previous change in this area). Sat May 16 12:08:17 GMT 2015 Olly Betts <olly@survex.com> * common/socket_utils.cc: Fix casts on the 4th argument of setsockopt() so they compile on platforms which expect char * or const char * there. Fri May 15 13:24:55 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: Enhance the probe for whether the test harness can use RTTI so that it works for xlC (which defaults to not generating RTTI). Fri May 15 12:28:32 GMT 2015 Olly Betts <olly@survex.com> * common/debuglog.h: Remove commented out STATIC_ASSERT() which is conceptually flawed. Fri May 15 12:27:45 GMT 2015 Olly Betts <olly@survex.com> * common/omassert.h,common/pack.h,tests/api_wrdb.cc, tests/internaltest.cc: Replace STATIC_ASSERT() with C++11's static_assert(). Fri May 15 11:34:47 GMT 2015 Olly Betts <olly@survex.com> * common/stringutils.cc: Use static_assert rather than an adhoc check via a negative array size. Fri May 15 11:16:18 GMT 2015 Olly Betts <olly@survex.com> * common/io_utils.cc: Replace the code used on platforms without F_DUPFD with simpler code which actually compiles. Fri May 15 03:43:16 GMT 2015 Olly Betts <olly@survex.com> * pkgconfig/xapian-core.pc.in: Include @ldflags@ in pkg-config .pc file. Fri May 15 03:33:40 GMT 2015 Olly Betts <olly@survex.com> * pkgconfig/xapian-core.pc.in: Fix include directory reported by pkg-config. Fri May 15 03:31:43 GMT 2015 Olly Betts <olly@survex.com> * configure.ac,pkgconfig/xapian-core.pc.in,xapian-config.in: When compiling with xlC on AIX, _LARGE_FILES gets defined by AC_SYS_LARGEFILE to enable large file support, and defining this changes the ABI of std::string, so include it in xapian-config --cxxflags and the pkg-config equivalent. Fri May 15 02:52:10 GMT 2015 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/perftest/Makefile.mk: Use $(NO_INSTALL) rather than @NO_INSTALL@ to allow make-time overriding. Fri May 15 02:33:11 GMT 2015 Olly Betts <olly@survex.com> * tests/soaktest/Makefile.mk: Use -no-fast-install instead of -no-install for linking soaktest on platforms where libtool issues a warning for -no-install, like we already do for all the other test programs. Fri May 15 02:14:16 GMT 2015 Olly Betts <olly@survex.com> * examples/simpleexpand.cc,examples/simpleindex.cc, examples/simplesearch.cc: '#include <config.h>' in the examples, as when compiling with xlC on AIX, _LARGE_FILES gets defined by AC_SYS_LARGEFILE to enable large file support, and defining this changes the ABI of std::string, so it also needs to be defined when compiling code using Xapian. Thu May 14 03:17:28 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.cc: Correct delete to delete []. Tue May 12 14:33:28 GMT 2015 Olly Betts <olly@survex.com> * docs/deprecation.rst: Add deprecation of --preserve-nonduplicates which happened in 1.2.4. Tue May 12 07:38:12 GMT 2015 Olly Betts <olly@survex.com> * Makefile.am: Actually use $(NO_UNDEFINED). Tue May 12 06:40:38 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: Only pass -no-undefined when linking the library if it's actually required for the current platform. Sun C++ doesn't link the C++ runtime libraries to shared objects, so the build fails with -no-undefined. Tue May 12 06:39:26 GMT 2015 Olly Betts <olly@survex.com> * INSTALL,configure.ac: Fix comment and doc typos. Tue May 12 03:49:38 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: Add workaround for odd <cmath> issue with Sun C++. Tue May 12 03:49:15 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: On Solaris, inet_ntop() needs -lnsl. Tue May 12 03:48:48 GMT 2015 Olly Betts <olly@survex.com> * api/keymaker.cc: Fix error compiling with Sun C++. Tue May 12 03:07:06 GMT 2015 Olly Betts <olly@survex.com> * include/xapian/attributes.h: Fix typo in definition of XAPIAN_NOTHROW() definition for C++11 compilers other than GCC and MSVC. Tue May 12 01:48:19 GMT 2015 Olly Betts <olly@survex.com> * exception_data.pm: Fix not to parse mentions of XAPIAN_THROW() in comments. Mon May 11 14:16:22 GMT 2015 Olly Betts <olly@survex.com> * HACKING: libtool 2.4.6 is now used for snapshots and releases. Mon May 11 14:14:04 GMT 2015 Olly Betts <olly@survex.com> * common/str.cc,expand/bo1eweight.cc,geospatial/geoencode.cc, matcher/multiandpostlist.cc,net/remoteconnection.cc,weight/: Fixes for errors when compiling with Sun C++. Mon May 11 10:53:47 GMT 2015 Olly Betts <olly@survex.com> * common/log2.h,configure.ac: Change probe for log2() to check for a declaration and '#include <cmath>' to get it to fix build on Solaris with Sun C++. C++11 compilers should all provide log2(), but let's not rely on that just yet as it's easy to provide a fallback implementation. Mon May 11 07:04:06 GMT 2015 Olly Betts <olly@survex.com> * common/getopt.cc,net/tcpserver.cc: Fix clang warnings on OS X. Reported by Germán M. Bravo. Mon May 11 06:29:42 GMT 2015 Olly Betts <olly@survex.com> * api/constinfo.cc,docs/doxygen_api.conf.in, docs/doxygen_source.conf.in,include/xapian/: C++11's noexcept needs to be on definitions as well as declarations, but GCC function attributes must only be on declarations, so split XAPIAN_NOEXCEPT off from XAPIAN_NOTHROW. Sun May 10 13:26:24 GMT 2015 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser_internal.h, tests/api_none.cc: Add support for optional reference counting of FieldProcessor objects. Sun May 10 11:07:36 GMT 2015 Olly Betts <olly@survex.com> * include/xapian/intrusive_ptr.h,include/xapian/queryparser.h, queryparser/queryparser.lemony,queryparser/queryparser_internal.h, tests/api_none.cc: Add support for optional reference counting of ValueRangeProcessor objects. Thu May 07 12:42:11 GMT 2015 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc,common/socket_utils.cc,common/socket_utils.h, net/replicatetcpclient.cc,net/replicatetcpclient.h: Set SO_KEEPALIVE for xapian-replicate's connection to the master, and add command line option to allow setting socket-level timeouts (SO_RCVTIMEO and SO_SNDTIMEO) on platforms that support them. Fixes #546, reported by nkvoll. Thu May 07 12:00:15 GMT 2015 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc: Fix connection timeout to be 10 seconds rather than 10000 seconds which is has been since 1.2.3. Tue May 05 02:38:37 GMT 2015 Olly Betts <olly@survex.com> * net/tcpserver.cc: Use inet_ntop() instead of inet_ntoa() - the latter isn't thread-safe. Tue May 05 02:05:27 GMT 2015 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Update documentation to reflect that the database passed to QueryParser::set_database() is no longer used to expand wildcards and partial terms. Mon May 04 11:32:40 GMT 2015 Olly Betts <olly@survex.com> * api/postlist.cc,api/postlist.h,common/remoteprotocol.h,matcher/, net/serialise.cc,tests/api_backend.cc: Fix sort by value when multiple databases are in use and one or more are remote. Fixes #674, reported by Dylan Griffith. Sun May 03 12:41:58 GMT 2015 Olly Betts <olly@survex.com> * docs/deprecation.rst,include/xapian/queryparser.h, queryparser/queryparser.cc,queryparser/queryparser.lemony, queryparser/queryparser_internal.h,tests/queryparsertest.cc: Allow setting the expansion limits and types for wildcards and partial terms in the QueryParser independently. Partial terms now default to the 100 most frequent matching terms. Deprecate set_max_wildcard_expansion() in favour of new set_max_expansion() method which can set expansion limits for partial terms too. Completes #608. Sat May 02 09:20:30 GMT 2015 Olly Betts <olly@survex.com> * api/query.cc,api/queryinternal.cc,api/queryinternal.h, include/xapian/query.h,queryparser/queryparser.lemony, tests/api_query.cc: Add a choice of ways to apply the limit on wildcard expansion. Fixes #608, reported by boomboo. Fri May 01 14:53:16 GMT 2015 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Reenable replicate3 for glass, as it no longer fails. Fri May 01 13:38:48 GMT 2015 Olly Betts <olly@survex.com> * common/autoptr.h: Just map AutoPtr to std::unique_ptr now we require C++11. Fri May 01 13:33:28 GMT 2015 Olly Betts <olly@survex.com> * matcher/multimatch.cc,net/progclient.cc: IRIX is dead, so remove lingering traces of support for SGI's compiler. Fri May 01 08:59:13 GMT 2015 Olly Betts <olly@survex.com> * HACKING,INSTALL: STLport was last released in 2008, so (a) it's no longer actively developed and (b) it won't support C++11, so strip out documentation which mentions it. * configure.ac: For Sun's C++ compiler, -std=c++11 enables C++11 support, and is incompatible with -library=stlport, so remove code to enable that later option. Fri May 01 01:37:24 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: Tweak comment to match xapian-omega's configure. Fri May 01 01:36:41 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: Set default value for AUTOM4TE before AC_OUTPUT so the default will actually get used. Fri May 01 01:36:05 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: Add explicit AC_CANONICAL_HOST. Fri May 01 01:08:56 GMT 2015 Olly Betts <olly@survex.com> * api/matchspy.cc,tests/api_nodb.cc: Eliminate needless uses of AutoPtr. Fri May 01 00:37:42 GMT 2015 Olly Betts <olly@survex.com> * tests/internaltest.cc: Remove self-assignment check of AutoPtr as that doesn't work under C++11 when AutoPtr is unique_ptr, and add other checks of reassignment which should work for both. Fri May 01 00:37:09 GMT 2015 Olly Betts <olly@survex.com> * tests/api_postingsource.cc: Remove C++11 specific test which doesn't actually work with C++11. Fri May 01 00:33:43 GMT 2015 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc, tests/harness/backendmanager_remotetcp.cc, tests/harness/testsuite.h,tests/harness/testutils.h: Add spaces between literal strings and macros which expand to literal strings for C++11 compatibility. Thu Apr 30 14:41:59 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc,backends/glass/glass_table.cc: Remove duplicate includes of "omassert.h". Thu Apr 30 14:41:31 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_freelist.cc: Add missing explicit include of "omassert.h", Thu Apr 30 14:38:06 GMT 2015 Olly Betts <olly@survex.com> * HACKING,backends/chert/chert_table.cc, backends/chert/chert_version.cc,backends/glass/glass_table.cc, common/omassert.h: Remove CompileTimeAssert() and just use C++11's static_assert instead. Thu Apr 30 14:28:10 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_freelist.cc: Drop conditionals for GCC >= 3.4 as we now require >= 4.4. Thu Apr 30 14:26:13 GMT 2015 Olly Betts <olly@survex.com> * api/omdatabase.cc,include/xapian/queryparser.h, tests/harness/backendmanager.h: Drop workarounds for ancient Sun C++ compilers, as we now require a recent version for C++11 support. Thu Apr 30 14:22:46 GMT 2015 Olly Betts <olly@survex.com> * INSTALL: IRIX is past EOL so drop information about IRIX make. Thu Apr 30 14:22:10 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: OS X >= 10.4 apparently supports symbol visibility, so update comment. Thu Apr 30 14:21:37 GMT 2015 Olly Betts <olly@survex.com> * common/unordered_map.h: No longer relevant now we require C++11. Thu Apr 30 13:59:16 GMT 2015 Olly Betts <olly@survex.com> * HACKING,INSTALL,configure.ac,xapian-config.in: Require a C++ compiler - the minimum required GCC version is now probably GCC 4.4. Thu Apr 30 13:48:54 GMT 2015 Olly Betts <olly@survex.com> * xapian-config.in: Remove superfluous comment left over from code refactoring. Thu Apr 30 13:47:30 GMT 2015 Olly Betts <olly@survex.com> * xapian-config.in: Fix the include directory reported by --swigflags for an installed copy to include the "/xapian-1.3" component for a development version. Thu Apr 30 12:34:14 GMT 2015 Olly Betts <olly@survex.com> * api/omenquire.cc: Fix for C++11 where AutoPtr is unique_ptr. Thu Apr 30 12:20:22 GMT 2015 Olly Betts <olly@survex.com> * api/replication.cc,backends/chert/chert_database.cc, backends/chert/chert_table.cc,backends/chert/chert_version.cc, backends/glass/,backends/remote/remote-database.cc,bin/, examples/copydatabase.cc,examples/quest.cc, examples/xapian-metadata.cc,net/tcpclient.cc, queryparser/queryparser.lemony: Add spaces between literal strings and macros which expand to literal strings for C++11 compatibility. Thu Apr 30 12:17:17 GMT 2015 Olly Betts <olly@survex.com> * bin/xapian-tcpsrv.cc: Use STRINGIZE() to include defaults literally in string for --help. Thu Apr 30 10:48:11 GMT 2015 Olly Betts <olly@survex.com> * HACKING: Update URL. Thu Apr 30 10:47:10 GMT 2015 Olly Betts <olly@survex.com> * m4/ax_cxx_compile_stdcxx_11.m4: Add support for HP's aCC (untested as I no longer have access to this compiler). Thu Apr 30 05:30:13 GMT 2015 Olly Betts <olly@survex.com> * m4/ax_cxx_compile_stdcxx_11.m4: Add macro from autoconf-archive to detect any compiler options needed to enable C++11 support (not yet used). Thu Apr 30 05:14:56 GMT 2015 Olly Betts <olly@survex.com> * common/posixy_wrapper.cc: O_RDONLY, etc are better described as "POSIX" than "ANSI". Thu Apr 30 05:12:33 GMT 2015 Olly Betts <olly@survex.com> * configure.ac,xapian-config.in: Rename ANSI_CXXFLAGS to ISO_CXXFLAGS and talk about ISO C++ in configure messages. Thu Apr 30 05:10:40 GMT 2015 Olly Betts <olly@survex.com> * .gitignore: Adjust for m4 subdirectory to only match files which aren't in git. Thu Apr 30 04:07:05 GMT 2015 Olly Betts <olly@survex.com> * xapian-config.in: Fix typo so cached result of test in is_uninstalled() is actually used on subsequent calls. Fixes #676, reported (with patch) by Ryan Schmidt. Wed Apr 29 03:00:32 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/glass/glass_database.cc: Fix get_changeset_revisions() not to potentially read uninitialised data if a changeset file is truncated. Wed Apr 29 00:09:57 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_databasereplicator.cc: Use return value of get_message_chunk() rather than ignoring it and repeating the check in the caller. Tue Apr 28 22:22:01 GMT 2015 Olly Betts <olly@survex.com> * bin/xapian-progsrv.cc: Fix comment typo. Tue Apr 28 04:51:33 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.cc,backends/chert/chert_table.cc, backends/glass/glass_cursor.cc,backends/glass/glass_table.cc: Potentially increment the cursor version on cancel() or when the database is reopened, and flag the current cursor version as used when a cursor is rebuilt. Fixes #675, reported by Germán M. Bravo. * tests/api_backend.cc: Add regression test cursorbug1 based on reproducer supplied by Germán. Thu Apr 23 05:25:48 GMT 2015 Olly Betts <olly@survex.com> * matcher/exactphrasepostlist.cc: When matching an exact phrase, if a term doesn't occur where we want, we can advance "term 0" using skip_to() based on where the term does next occur rather than just calling next() on it. Thu Apr 23 05:03:28 GMT 2015 Olly Betts <olly@survex.com> * matcher/nearpostlist.cc: Be lazier about opening PositionList objects, as that causes a read of all the data for that positionlist. Thu Apr 23 04:11:30 GMT 2015 Olly Betts <olly@survex.com> * matcher/nearpostlist.cc: Make code for scanning window for duplicate positions a little clearer. Wed Apr 22 13:52:55 GMT 2015 Olly Betts <olly@tartarus.org> * matcher/nearpostlist.cc: Fix "/*" within comment. Wed Apr 22 12:53:37 GMT 2015 Olly Betts <olly@survex.com> * matcher/nearpostlist.cc,matcher/nearpostlist.h: Reimplement OP_NEAR check. The new implementation consistently requires the terms to occur at different positions, and fixes some false negatives. * tests/api_query.cc: Add regression tests as loosenear1. Mon Apr 20 12:53:04 GMT 2015 Olly Betts <olly@survex.com> * matcher/phrasepostlist.cc,matcher/phrasepostlist.h: Reimplement OP_PHRASE check for loose phrases - the existing implementation was buggy, giving both false positives and false negatives in rare cases. Fixes #653, reported by Jean-Francois Dockes. * tests/api_query.cc: Add regression test loosephrase1. Thu Apr 16 06:51:31 GMT 2015 Olly Betts <olly@survex.com> * common/errno_to_string.cc: Fix to build with Sun's C++ compiler. Thu Apr 16 06:43:40 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: Fix XAPIAN_TEST_CXXFLAGS macro to append to the existing value of the variable, not its name. Wed Apr 15 11:36:40 GMT 2015 Olly Betts <olly@tartarus.org> * tests/api_wrdb.cc: Just #ifdef out the problematic part of the adddoc5 testcase for clang. Wed Apr 15 06:25:13 GMT 2015 Olly Betts <olly@survex.com> * common/serialise-double.cc: Fix bug in recent fix. * tests/unittest.cc: Extend serialise_double() to provide regression test. Thu Apr 09 00:40:24 GMT 2015 Olly Betts <olly@survex.com> * net/remoteconnection.h: Avoid compiler warning from clang in new getaddrinfo() code. Wed Apr 08 11:59:00 GMT 2015 Olly Betts <olly@survex.com> * common/serialise-double.cc: Add missing cast to unsigned char when we check if the value will fit in the double type. On machines with IEEE-754 doubles (which is most current platforms), which change makes no difference as we're checking if the value cast to a double is > 255.<something> which is never the case. Wed Apr 08 10:42:43 GMT 2015 Olly Betts <olly@survex.com> * common/serialise-double.cc: Avoid reading one byte past the end of the serialised value. In practice this was harmless on most platforms, as dbl_max_mantissa is 255 for IEEE-754 format doubles, and at least GCC's std::string keeps the buffer nul-terminated. Reported by Germán M. Bravo in github PR#67. * tests/unittest.cc: Copy the serialised value to a temporary buffer before calling unserialise_double() so that valgrind can detect if we read one byte past the end. Tue Apr 07 23:02:21 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/: Fix problems with get_unique_terms() on a modified chert database. * tests/: Add more test coverage for get_unique_terms(). Mon Apr 06 00:22:22 GMT 2015 Olly Betts <olly@survex.com> * tests/harness/backendmanager_multi.cc, tests/harness/backendmanager_remote.cc: Cleaner code for checking the backend subtype. Mon Mar 30 03:22:29 GMT 2015 Olly Betts <olly@survex.com> * backends/flint_lock.cc: If available, use F_DUPFD to dup to a file descriptor which is >= 2. Mon Mar 30 01:16:06 GMT 2015 Olly Betts <olly@survex.com> * api/error.cc,common/safenetdb.h,configure.ac,generate-exceptions, net/remoteconnection.h,net/tcpclient.cc,net/tcpserver.cc: Use getaddrinfo() instead of gethostbyname(), as a step towards IPv6 support (currently we still only look for IPv4 addresses). See #374. Sun Mar 29 03:56:12 GMT 2015 Olly Betts <olly@survex.com> * Makefile.am,docs/Makefile.am: Add missing quoting. Mon Mar 23 06:33:39 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc,backends/glass/glass_table.cc, common/io_utils.cc,common/io_utils.h,tests/api_backend.cc: Avoid using fds < 3 for writable database tables, as it risks corruption is some code in the same process tries to write to stdout or stderr without realising it is closed. (Partly addresses #651) Sat Mar 21 22:11:53 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_cursor.cc: Add comment warning that B can't be used in GlassCursor's destructor. Tue Mar 17 03:51:00 GMT 2015 Olly Betts <olly@survex.com> * matcher/exactphrasepostlist.cc: Fixed reversed conditonal for picking the shorter position list for an exact phrase of two terms. The difference this fix makes isn't dramatic, but can be measured (at least with cachegrind). Thanks to kbwt for spotting this. Fri Mar 13 11:21:37 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_compact.cc: Make sure we open all the tables of a database at the same revision when compacting. (Fixes #649) Thu Mar 12 04:09:07 GMT 2015 Olly Betts <olly@survex.com> * examples/xapian-metadata.cc: Add 'list' subcommand to list all the metadata keys. Mon Mar 09 03:08:14 GMT 2015 Olly Betts <olly@survex.com> * include/xapian/constants.h,include/xapian/queryparser.h: Doxygen comment wording tweaks. Mon Mar 09 03:07:36 GMT 2015 Olly Betts <olly@survex.com> * include/xapian/constants.h,include/xapian/queryparser.h, include/xapian/weight.h: Doxygen comment formatting tweaks. Sat Mar 07 11:27:23 GMT 2015 Olly Betts <olly@tartarus.org> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Extend the set of characters allowed in the start of a range to be anything except for '(' and characters <= ' '. This better matches what's accepted for a range end (anything except for ')' and characters <= ' '). Reported by Jani Nikula. Thu Mar 05 11:13:56 GMT 2015 Olly Betts <olly@survex.com> * backends/dbcheck.cc: When checking a single table, handle the ".glass" extension on glass database tables, and use the extension to guide the decision of which backend the table is from. Thu Mar 05 04:38:21 GMT 2015 Olly Betts <olly@survex.com> * backends/dbcheck.cc: Don't hard-code string length as number. Mon Feb 23 04:48:24 GMT 2015 Olly Betts <olly@survex.com> * api/emptypostlist.cc,api/emptypostlist.h: Fix handling of OP_WILDCARD which expands to no terms when used under OP_SYNONYM. * tests/api_query.cc: Regression test dualprefixwildcard1. Mon Feb 23 04:26:52 GMT 2015 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Fix compilation issue with previous change. Thu Feb 19 08:04:17 GMT 2015 Olly Betts <olly@survex.com> * tests/api_anydb.cc,tests/api_nodb.cc,tests/harness/testrunner.cc, tests/harness/testsuite.cc,tests/stemtest.cc: Use std::string's assign() or construct from substring rather than assigning or initialising with the result of calling substr(). Tue Feb 17 18:31:25 GMT 2015 Olly Betts <olly@survex.com> * include/Makefile.mk: Ship new file from previous commit. Tue Feb 17 05:14:02 GMT 2015 Olly Betts <olly@survex.com> * api/Makefile.mk,api/constinfo.cc,api/version.cc,include/xapian.h, include/xapian/constinfo.h,include/xapian/stem.h,languages/stem.cc: Put the list of stemmers into the vinfo struct and renamed it to const_info. Mon Feb 16 10:18:51 GMT 2015 Olly Betts <olly@survex.com> * configure.ac,languages/: Merge generate-allsnowballheaders script into collate-sbl. Mon Feb 16 04:15:46 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_check.cc: Fix xapian-check on a single table. Fri Feb 13 10:18:45 GMT 2015 Olly Betts <olly@survex.com> * generate-exceptions: Note that the change to not negate WSAGetLastError() values internally was backported for 1.2.20. Fri Feb 13 05:57:13 GMT 2015 Olly Betts <olly@survex.com> * api/,common/remoteprotocol.h,exception_data.pm, include/xapian/query.h,include/xapian/queryparser.h,matcher/, queryparser/queryparser.lemony,tests/api_qpbackend.cc, tests/queryparsertest.cc,weight/weightinternal.cc, weight/weightinternal.h: Add new OP_WILDCARD query operator, which expands wildcards lazily, so now we create the PostList tree for a wildcard directly, rather than creating an intermediate Query tree. See ticket#48. Wed Feb 11 22:45:26 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_table.cc: mid_point() should never fail to find a split point, but just in case it does, return a usable answer. Wed Feb 11 22:44:35 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc: mid_point() should never fail to find a split point, but just in case it does, return a usable answer. Mon Feb 09 02:56:54 GMT 2015 Olly Betts <olly@survex.com> * api/queryinternal.cc,matcher/localsubmatch.cc, matcher/localsubmatch.h,matcher/queryoptimiser.h,tests/api_query.cc: Don't convert an unweighted term which indexes all docs to an all docs postlist if we need positional data for it. Sun Jan 25 23:48:34 GMT 2015 Olly Betts <olly@survex.com> * configure.ac: Check for declaration of fdatasync() first, as OS X has a dummy implementation in the library which is not prototyped in any header. Reported by Vlad Shablinsky. Fri Jan 09 05:19:20 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_freelist.cc: Fix freelist handling to allow for the newly loaded first block of the freelist being already used up. Fri Jan 09 04:38:03 GMT 2015 Olly Betts <olly@survex.com> * common/debuglog.cc: Add '#include "errno_to_string.h"'. Thu Jan 08 05:59:01 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_compact.cc: Fix compaction of position tables for the change in key format. Tue Jan 06 02:22:37 GMT 2015 Olly Betts <olly@survex.com> * HACKING: Now using automake 1.15 and libtool 2.4.4. Mon Jan 05 08:39:47 GMT 2015 Olly Betts <olly@survex.com> * Makefile.am: Remove the generated .pc file on "make distclean". Mon Jan 05 04:43:35 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_table.cc: In enter_key(), we know the insertion point will be the position after the current cursor position, so there's no need to call find_in_block() to locate it. Mon Jan 05 04:42:17 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_table.cc: More assertions that c is valid. Mon Jan 05 01:52:54 GMT 2015 Olly Betts <olly@survex.com> * backends/chert/chert_dbcheck.cc,backends/glass/glass_dbcheck.cc: Cross-check the position and postlist tables and report positional data for non-existent documents. Mon Jan 05 01:28:28 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_dbcheck.cc: Fix decoding of positionlist keys - this hadn't been updated for the key format change. Mon Jan 05 01:25:17 GMT 2015 Olly Betts <olly@survex.com> * backends/glass/glass_check.cc: Check that dir_end is odd. Tue Dec 23 21:26:49 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc: More assertions about c. Tue Dec 23 04:26:38 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc: Fix assertion to allow for c being DIR_START - D2 in leaf blocks. Tue Dec 23 04:01:54 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc: Add more assertions that c has suitable values. Tue Dec 23 03:58:00 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc: In enter_key(), we know the insertion point will be the position after the current cursor position, so there's no need to call find_in_block() to locate it. Tue Dec 23 03:11:12 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_check.cc: Add check that dir_end is odd. Mon Dec 22 23:28:54 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc: Document situations in which the cursor can be left pointing before DIR_START. Mon Dec 22 03:39:14 GMT 2014 Olly Betts <olly@survex.com> * backends/glass/glass_freelist.cc: Add more freelist assertions. Mon Dec 22 03:15:50 GMT 2014 Olly Betts <olly@survex.com> * backends/glass/glass_freelist.cc: Allow restricting the number of freelist entries per block by setting GLASS_FREELIST_SIZE. Sun Dec 21 20:55:47 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/weight.h: Document the enum stat_flags values. Sun Dec 21 20:51:47 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/weight.h: Fix indentation. Sat Dec 20 10:40:44 GMT 2014 Sébastien Debrard <sebastien.debrard@gmail.com> * Makefile.am,configure.ac,pkgconfig/,xapian-core.spec.in: Generate a file for pkg-config. (Fixes#540) Fri Dec 19 22:30:39 GMT 2014 Olly Betts <olly@survex.com> * common/errno_to_string.cc: Need stdlib.h under mingw. Fri Dec 19 04:38:56 GMT 2014 Olly Betts <olly@survex.com> * backends/glass/glass_check.cc: Distinguish between a block in use and in the freelist, and a block in the freelist more than once. Fri Dec 19 04:37:14 GMT 2014 Olly Betts <olly@survex.com> * backends/glass/glass_freelist.cc,backends/glass/glass_freelist.h: Avoid any mutual recursion between GlassFreeList::get_block() and GlassFreeList::mark_block_unused(). Thu Dec 18 10:28:17 GMT 2014 Olly Betts <olly@survex.com> * common/safesyssocket.h: Include safeerrno.h when we use errno. Thu Dec 18 06:17:18 GMT 2014 Olly Betts <olly@survex.com> * net/tcpclient.cc: Use POSIX O_NONBLOCK in preference to O_NDELAY which has different semantics on BSD and System V. Wed Dec 17 02:59:54 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Drop probe for gethostbyaddr() - we've not used it for over 8.5 years (since e33972a5). Wed Dec 17 02:33:36 GMT 2014 Olly Betts <olly@survex.com> * common/internaltypes.h: Take CHAR_BIT into account when choosing types for uint2, uint4 and uint8. Wed Dec 17 02:30:14 GMT 2014 Olly Betts <olly@survex.com> * api/error.cc,backends/chert/chert_btreebase.cc, backends/chert/chert_table.cc,backends/flint_lock.cc, backends/glass/glass_table.cc,common/,configure.ac, net/remoteconnection.cc: Add errno_to_string() function which is thread-safe where a suitable alternative to strerror() exists. Tue Dec 16 06:54:32 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Add missing m4 quoting. Generated configure file is unchanged. Tue Dec 16 06:28:21 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Don't check for strerror - it's specified by C89, and we weren't using the result of the check anyway. Tue Dec 16 03:59:39 UTC 2014 Sébastien Debrard <sebastien.debrard@gmail.com> * api/omenquire.cc,backends/multi/multi_postlist.cc, tests/api_anydb.cc,tests/api_db.cc: Prefer ++x to x++. Tue Dec 16 03:50:41 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Define MINGW_HAS_SECURE_API under mingw to get _putenv_s() declared in stdlib.h. Sun Dec 14 09:42:01 GMT 2014 Olly Betts <olly@survex.com> * api/omdatabase.cc,include/xapian/database.h: Make overloaded forms of Database::allterms_begin() and Database::allterms_end() use default parameters instead. Sat Dec 13 02:48:10 GMT 2014 Olly Betts <olly@survex.com> * api/error.cc,generate-exceptions,net/remoteconnection.h: On Windows, avoid defining EADDRINUSE, etc if they're already defined, and use WSAE* constants un-negated, as they start from a high value so won't collide with E* constants. Fri Dec 12 03:41:02 GMT 2014 Olly Betts <olly@survex.com> * backends/glass/glass_freelist.cc: Avoid infinite recursion when we hit the end of the freelist block we're reading and the end of the block we're writing at the same time. Thu Dec 11 20:45:28 GMT 2014 Olly Betts <olly@survex.com> * backends/glass/glass_version.cc: Use str() rather than sprintf(). Thu Dec 11 03:21:46 GMT 2014 Olly Betts <olly@survex.com> * net/tcpserver.cc: Fix comment. Thu Dec 11 03:20:56 GMT 2014 Olly Betts <olly@survex.com> * api/compactor.cc,api/error.cc,api/replication.cc, backends/chert/chert_databasereplicator.cc, backends/chert/chert_table.cc, backends/glass/glass_databasereplicator.cc, backends/glass/glass_table.cc: Remove unnecessary includes of <cstdio> in files which used to need it for rename() but no longer do. Fri Dec 05 11:30:05 GMT 2014 Olly Betts <olly@survex.com> * ChangeLog,ChangeLog.0,ChangeLog.examples, backends/chert/chert_btreebase.cc,backends/chert/chert_database.h, languages/compiler/driver.c,languages/turkish.sbl: Fix typos reported by Veres Lajos. Thu Dec 04 02:17:46 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Fix detection of fdatasync(), which appears to have been broken practically forever - this means we've probably been using fsync() instead, which probably isn't a big additional overhead. Wed Dec 03 09:38:24 GMT 2014 Olly Betts <olly@survex.com> * docs/Makefile.am: Quote png filenames we run through pngcrush to avoid problems if there's an oddly-named extra PNG file in the directory. Tue Dec 02 02:58:59 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_database.h, backends/chert/chert_databasereplicator.cc,backends/, backends/glass/glass_compact.cc,backends/glass/glass_database.cc, backends/glass/glass_databasereplicator.cc, backends/remote/remote-database.cc,backends/remote/remote-database.h, include/xapian/constants.h,include/xapian/database.h, include/xapian/dbfactory.h,net/,tests/api_backend.cc: Add new flag Xapian::DB_RETRY_LOCK which allows opening a database for writing to wait until it can get a write lock. (fixes#275) Tue Dec 02 02:55:33 GMT 2014 Olly Betts <olly@survex.com> * backends/glass/glass_database.h: Document parameter to internal method Sun Nov 30 15:55:20 GMT 2014 James Aylett <james@tartarus.org> * HACKING: improve OS X docs build information. Thu Nov 27 04:00:38 GMT 2014 Olly Betts <olly@survex.com> * backends/dbcheck.cc,backends/flint_lock.cc,backends/flint_lock.h, examples/copydatabase.cc,tests/api_backend.cc: Drop support code for building on OS/2 with EMX - EMX was last updated in 2001 and comes with GCC 3.2.1, which pre-dates C++11 and we're going to require a compiler with good C++11 support in the near future. Thu Nov 27 03:15:51 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Rename svn-ci to xapian-commit (see #621). Thu Nov 27 03:05:18 GMT 2014 Olly Betts <olly@survex.com> * backends/flint_lock.cc: Use F_OFD_SETLK where available, which avoids having to fork() a child process to hold the lock. This currently requires Linux kernel >= 3.15, but it has been submitted to POSIX so hopefully will be widely supported eventually. Thanks to Austin Clements for pointing out this now exists. Wed Nov 26 04:03:45 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/geospatial.h: Drop documentation of parameter which the method doesn't actually take. Wed Nov 26 04:03:24 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Fix incorrect parameter name in doxygen comment. Wed Nov 26 03:29:28 GMT 2014 Olly Betts <olly@survex.com> * docs/doxygen_source.conf.in: @MAKEINDEX@ isn't substituted, so just use literal 'makeindex'. Wed Nov 26 03:28:32 GMT 2014 Olly Betts <olly@survex.com> * docs/doxygen_api.conf.in,docs/doxygen_source.conf.in: Update to avoid warnings about obsolete tags from newer doxygen. Tue Nov 25 21:13:51 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Update link to cygwin FAQ. Tue Nov 25 01:55:17 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Update details of building Xapian packages. Tue Nov 25 01:48:44 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Update for SVN to git migration. Mon Nov 24 19:50:32 GMT 2014 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update for 1.3.2. Mon Nov 24 03:59:02 GMT 2014 Olly Betts <olly@survex.com> * languages/Makefile.mk: Add missing '\' at end of line in previous commit. Mon Nov 24 03:14:22 GMT 2014 Olly Betts <olly@survex.com> * languages/: Install the stopword lists as ${pkgdatadir}/stopwords/LANG.list. Sun Nov 23 23:56:02 GMT 2014 Olly Betts <olly@survex.com> * HACKING,INSTALL,NEWS,NEWS.SKELETON,api/compactor.cc, api/replication.cc,backends/Makefile.mk,backends/brass/, backends/databasereplicator.cc,backends/dbcheck.cc, backends/dbfactory.cc,backends/glass/,bin/Makefile.mk, common/Makefile.mk,common/pack.h,common/pretty.h,configure.ac,docs/, include/xapian/,net/replication_protocol.rst,tests/,tests/harness/: Rename 'brass' backend to 'glass'. Sun Nov 23 22:23:00 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_freelist.cc,backends/brass/brass_freelist.h: Set the revision in BrassFreeList::write_block(). Thu Nov 20 21:36:35 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_freelist.h: Reset flw_appending when we unpack a freelist. Thu Nov 20 21:36:15 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_freelist.cc: Fix comment typo. Thu Nov 20 03:53:32 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_freelist.cc: Fix end-of-block condition when walking the freelist during database consistency checking. Thu Nov 20 02:32:37 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_freelist.cc: Wrap comment line. Thu Nov 20 02:31:54 GMT 2014 Olly Betts <olly@survex.com> * matcher/exactphrasepostlist.cc: Straighten out confusing comments. Sun Nov 16 23:49:38 GMT 2014 Olly Betts <olly@survex.com> * NEWS: Update. Thu Nov 13 20:35:25 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_freelist.cc: Wrap comment line. Tue Nov 11 22:17:48 GMT 2014 Olly Betts <olly@survex.com> * api/queryinternal.cc,api/queryinternal.h: Move all the get_op() definitions out of the header as they're virtual and there's not much scope for devirtualisation. Tue Nov 11 02:58:05 GMT 2014 Olly Betts <olly@survex.com> * exception_data.pm: Just have one copy of the licence boilerplate. Sun Nov 09 22:38:04 GMT 2014 Olly Betts <olly@survex.com> * NEWS: Update. Fri Nov 07 00:49:22 GMT 2014 Olly Betts <olly@survex.com> * queryparser/lemon.c: Sync change from upstream: "Modify the %nonassoc directive in lemon so that it generates a run-time error rather than a parsing conflict. This changes is due to a bug report on the mailing list. SQLite does not use the %nonassoc directive in its grammar so this change does not affect SQLite." Generated code is identical. Fri Nov 07 00:44:57 GMT 2014 Olly Betts <olly@survex.com> * queryparser/lemon.c: Sync change from upstream: "Fix harmless compiler warning in LEMON." Generated code is identical. Fri Nov 07 00:39:44 GMT 2014 Olly Betts <olly@survex.com> * queryparser/lemon.c: Sync change from upstream: "In LEMON, limit the size of the grammar file to 100MB. This ensures that the program will never experience integer overflow. To be doubly sure, use calloc() instead of malloc() when allocating arrays." Generated code is identical. Fri Nov 07 00:34:26 GMT 2014 Olly Betts <olly@survex.com> * queryparser/lemon.c: Sync change from upstream: "Add the "%token_class" directive to the LEMON parser generator. This opens up the possibility of simplifying the parser. Also remove all calls to sprintf(), strcpy(), and strcat() from LEMON to avoid compiler warnings on OpenBSD. (Aside: It is this change to avoid harmless compiler warnings that was the cause of the reason spat of bugs.)" Generated code is identical aside from whitespace changes in queryparser_token.h. Fri Nov 07 00:17:18 GMT 2014 Olly Betts <olly@survex.com> * queryparser/lemon.c: Sync change from upstream: "In Lemon, when comparing the output to the *.h file to see if it has changed, make sure that the proposed new output and the preexisting output are the same size before deciding that they are the same". Generated code is identical, but this will fix misbehaviour in a build from a non-clean tree if new tokens get appended to queryparser_token.h. Thu Nov 06 22:23:27 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Now using doxygen 1.8.8. Thu Nov 06 04:18:32 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Now using libtool 2.4.3. Wed Nov 05 21:38:45 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_check.cc: Handle DBCHECK_FIX for a zero sized .DB file. Wed Nov 05 21:02:21 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_check.cc: After we successfully regenerate baseA, remove any empty baseB file to prevent it causing problems. Tracked down with help from Phil Hands. Wed Nov 05 20:58:27 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_check.cc: Fix DBCHECK_FIX to be able to find a level 0 root block (happens when a table has sufficiently little data in that it all fits in one block). Sun Nov 02 21:14:43 GMT 2014 Olly Betts <olly@survex.com> * docs/deprecation.rst: Fix typo. Tue Oct 28 23:35:10 GMT 2014 Olly Betts <olly@survex.com> * docs/replication.rst: Fix typo. Tue Oct 28 02:34:54 GMT 2014 Olly Betts <olly@survex.com> * docs/postingsource.rst: Use a modern class in postingsource example. (Noted by James Aylett) Mon Oct 27 21:06:55 GMT 2014 Olly Betts <olly@survex.com> * docs/,net/Makefile.mk,net/remote_protocol.rst, net/replication_protocol.rst: Move the protocol docs for the remote and replication protocols into the net/ subdirectory. Mon Oct 27 19:19:59 GMT 2014 Olly Betts <olly@survex.com> * Makefile.am,api/Makefile.mk,api/dir_contents,backends/Makefile.mk, backends/brass/Makefile.mk,backends/brass/dir_contents, backends/chert/Makefile.mk,backends/chert/dir_contents, backends/dir_contents,backends/inmemory/Makefile.mk, backends/inmemory/dir_contents,backends/multi/Makefile.mk, backends/multi/dir_contents,backends/remote/Makefile.mk, backends/remote/dir_contents,bin/Makefile.mk,bin/dir_contents, common/Makefile.mk,common/dir_contents,configure.ac,dir_contents, docs/,examples/Makefile.mk,examples/dir_contents,expand/Makefile.mk, expand/dir_contents,geospatial/Makefile.mk,geospatial/dir_contents, include/Makefile.mk,include/dir_contents,include/xapian/dir_contents, languages/Makefile.mk,languages/dir_contents,m4-macros/dir_contents, m4/dir_contents,matcher/Makefile.mk,matcher/dir_contents, net/Makefile.mk,net/dir_contents,preautoreconf, queryparser/Makefile.mk,queryparser/dir_contents,tests/Makefile.am, tests/dir_contents,tests/harness/Makefile.mk, tests/harness/dir_contents,tests/perftest/Makefile.mk, tests/perftest/dir_contents,tests/soaktest/Makefile.mk, tests/soaktest/dir_contents,unicode/Makefile.mk,unicode/dir_contents, weight/Makefile.mk,weight/dir_contents: Remove the dir_contents files and all the machinery to handle them. Sat Oct 25 19:25:06 GMT 2014 Olly Betts <olly@survex.com> * api/query.cc,api/queryinternal.cc,api/smallvector.h, include/xapian/query.h: Stop using a reference where we may end up passing *NULL, as that's invalid. Thanks Nick Lewycky and ubsan for helping track this down. * tests/api_postingsource.cc: Fix testsuite to avoid passing literal NULL - pass nullptr for C++11 and don't bother with that part of the testcase for older C++. Sat Oct 25 19:05:07 GMT 2014 Olly Betts <olly@survex.com> * weight/dlhweight.cc,weight/dphweight.cc: Avoid dividing by zero when the collection size is 0. Fri Oct 24 23:04:24 GMT 2014 Gaurav Arora <gauravarora.daiict@gmail.com> * api/registry.cc,include/xapian/weight.h,tests/api_backend.cc, tests/api_nodb.cc,tests/api_weight.cc,weight/Makefile.mk, weight/lmweight.cc: Add Unigram LMWeight class. Tue Oct 21 19:19:32 GMT 2014 Olly Betts <olly@survex.com> * NEWS: Update for 1.2.19 release. Mon Oct 20 00:47:32 GMT 2014 Olly Betts <olly@survex.com> * .gitignore,docs/.gitignore,tests/.gitignore,unicode/.gitignore: Update to ignore new generated files. Sat Oct 18 06:15:52 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Don't provide a default expansion scheme name in Enquire::set_expansion_scheme(). Fri Oct 17 00:33:55 GMT 2014 Olly Betts <olly@survex.com> * api/queryinternal.cc,matcher/: Split NearPostList out into separate source file and header. Wed Oct 15 04:33:31 GMT 2014 Olly Betts <olly@survex.com> * api/queryinternal.cc,matcher/exactphrasepostlist.cc: We never try to build an ExactPhrasePostList over 0 or 1 terms, so simply assert we don't, rather than having an explicit check for it in the code. Tue Oct 14 02:47:02 GMT 2014 Olly Betts <olly@survex.com> * matcher/phrasepostlist.cc,tests/api_backend.cc: If the anchor term for a phrase is the first or last, set the min or max to its position respectively. Fixes the testcase in #657, but may not be a complete fix. Sun Oct 12 12:03:10 GMT 2014 Olly Betts <olly@survex.com> * NEWS: Update. Fri Oct 10 10:53:38 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_btreebase.cc,backends/chert/chert_table.cc: The table is sync-ed after the base file, so it's the last table we want to full-sync, not the last base file. Thu Oct 09 02:39:00 GMT 2014 Olly Betts <olly@survex.com> * common/debuglog.h: Suppress unused typedef warnings from debugging logging macros, which occur in functions which always exit via throwing an exception. Thu Oct 09 01:45:00 GMT 2014 Olly Betts <olly@survex.com> * tests/harness/fdtracker.cc: Fix code typo in recent change to skip apparent leaks of /dev/urandom. Thu Oct 09 01:43:02 GMT 2014 Olly Betts <olly@survex.com> * common/pretty.h: Fix debug logging code to compile with clang. (fixes #657, reported by Germán M. Bravo) Wed Oct 08 04:43:05 GMT 2014 Olly Betts <olly@survex.com> * matcher/multimatch.cc: If an empty RSet is specified, just handle it as if no RSet were specified. Mon Oct 06 21:41:58 GMT 2014 Olly Betts <olly@survex.com> * tests/Makefile.am: Remove .replicatmp upon "make clean". Fri Sep 19 05:31:15 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Fix compilation with clang. Reported by Germán M. Bravo. Fri Sep 12 17:54:34 GMT 2014 Olly Betts <olly@survex.com> * docs/overview.rst: Add missing database path to example of using xapian-progsrv in a stub database file. Thu Sep 11 21:33:49 GMT 2014 Olly Betts <olly@survex.com> * include/Makefile.mk,include/xapian/enquire.h, include/xapian/iterator.h: Add simple API to help with creating language-idiomatic iterator wrappers. Wed Sep 10 17:02:18 GMT 2014 Olly Betts <olly@survex.com> * api/queryinternal.cc: If we hit an unknown query serialisation, report the numeric code for it in the exception thrown. Wed Sep 10 16:59:20 GMT 2014 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Enable some disabled cases which actually work (in some cases with slightly tweaked expected answers which are equivalent to those that were shown). Fri Sep 05 17:41:53 GMT 2014 Olly Betts <olly@survex.com> * tests/harness/fdtracker.cc: Don't report apparent leaks of fds opened on /dev/urandom - at least on Linux, something in the C library seems to lazily open it, and the report of a possible leak followed by assurance that it's OK really is just noise we could do without. Fri Sep 05 14:06:08 GMT 2014 Olly Betts <olly@survex.com> * common/pretty.h: Add new brass-related types. Thu Sep 04 20:51:16 GMT 2014 Olly Betts <olly@survex.com> * backends/dbfactory.cc: Fix incorrect return type in logging annotations for InMemory::open(). Thu Sep 04 20:48:37 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc,backends/chert/chert_database.cc: Fix logging annotations for get_freqs() to reflect that it now has void return type. Thu Sep 04 20:21:32 GMT 2014 Olly Betts <olly@survex.com> * api/,backends/brass/,backends/chert/,backends/dbfactory_remote.cc, matcher/mergepostlist.cc,matcher/remotesubmatch.cc, net/progclient.cc: Add missing RETURN() markup for debug logging, highlighted by warning from recent GCC. Thu Sep 04 18:02:38 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_freelist.cc,backends/brass/brass_freelist.h: Make the BrassTable parameter a const pointer. Thu Sep 04 17:13:57 GMT 2014 Olly Betts <olly@survex.com> * common/io_utils.cc: Reading or writing a block will almost always complete in one call, so streamline that case. Wed Sep 03 19:30:00 GMT 2014 Olly Betts <olly@survex.com> * common/internaltypes.h: Update comment to reflect C++11 now being out. Wed Sep 03 19:29:09 GMT 2014 Olly Betts <olly@survex.com> * common/internaltypes.h,configure.ac: Make uint2 available. Wed Sep 03 19:27:41 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_database.h: Remove method which is declared but never defined. Wed Sep 03 18:52:40 GMT 2014 Olly Betts <olly@survex.com> * generate-exceptions,include/xapian/: Use XAPIAN_IN_XAPIAN_H rather than XAPIAN_INCLUDED_XAPIAN_H in the direct inclusion of subheader checks. Currently neither will catch the case of the user including a sub-header after <xapian.h> since the include guard for the subheader will stop us getting to the direct inclusion check, but XAPIAN_IN_XAPIAN_H will work if we move the check outside the include guard. Wed Sep 03 18:29:41 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/constants.h: Hide internal constants from external users. Tue Sep 02 14:15:50 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_spellingwordslist.h: Explicit include brass_cursor.h since we actually use BrassCursor here. Mon Sep 01 22:03:11 GMT 2014 Olly Betts <olly@survex.com> * tests/api_backend.cc: New testcase readonlyparentdir1 to ensure that commit works with a read-only parent directory. Mon Sep 01 21:51:45 GMT 2014 Olly Betts <olly@survex.com> * tests/harness/testrunner.h: Drop "flint" flag - flint was removed from trunk some time ago. Mon Sep 01 21:42:31 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/: Eliminate brass_doclen_t in favour of just using Xapian::termcount. Mon Sep 01 21:28:39 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_check.h: Tweak to ease upcoming big merge. Mon Sep 01 20:59:40 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/,backends/dbcheck.cc: Only create a btree entry for the document data if it is non-empty. The table is now lazily created, so won't exist if no documents have document data set. Mon Sep 01 16:54:44 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_record.cc,backends/brass/brass_record.h: Remove BrassRecordTable::get_doccount() as it is no longer used. Mon Sep 01 16:51:04 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/: Track doccount in the database stats. Mon Sep 01 16:27:15 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc: Update comment and exception message not to talk about "consistent state", as that was to do with the base files. Mon Sep 01 14:57:02 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc: Explicitly include brass_defs.h. Mon Sep 01 14:48:01 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/: Eliminate backends/brass/brass_types.h. Fri Aug 29 21:06:22 GMT 2014 Olly Betts <olly@survex.com> * backends/dbcheck.cc: Fix checking a single brass table. Thu Aug 28 16:04:59 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_check.cc,backends/chert/chert_check.cc, bin/xapian-check.cc,include/xapian/constants.h: Rename DBCHECK_SHOW_BITMAP to DBCHECK_SHOW_FREELIST. Tue Aug 26 22:31:32 GMT 2014 Olly Betts <olly@survex.com> * NEWS: Update. Fri Aug 22 09:09:52 GMT 2014 Gaurav Arora <gauravarora.daiict@gmail.com> * api/,backends/brass/,backends/chert/, backends/contiguousalldocspostlist.cc, backends/contiguousalldocspostlist.h,backends/database.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/multi/multi_postlist.cc,backends/multi/multi_postlist.h, backends/remote/,common/remoteprotocol.h,docs/sorting.rst, include/xapian/database.h,include/xapian/postingiterator.h, include/xapian/weight.h,matcher/,net/remoteserver.cc, net/remoteserver.h,tests/,weight/: Provide the number of distinct terms in the document to weighting schemes. The current implementation for the disk-based backends is a bit inefficient as it has to look at the termlist table entry for the document. Thu Aug 21 12:29:57 GMT 2014 Olly Betts <olly@survex.com> * weight/weight.cc: Add missing parameter to debug logging for Weight::init_(). Tue Aug 19 14:25:39 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_table.h: Don't sync handles for tables which aren't open. Reported by Austin Clements. Fri Aug 15 13:13:43 GMT 2014 Olly Betts <olly@survex.com> * HACKING: freecode no longer accepts updates, so drop that item from the release checklist. Wed Aug 13 23:55:24 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_btreebase.cc,backends/chert/chert_table.cc: Only full sync the final base file written. Mon Aug 11 05:30:20 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: For Sun's C++ compiler, pass -library=Crun separately since libtool looks for " -library=stlport4 " (with the spaces). (fixes#650) Mon Aug 11 03:00:15 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Rename XAPIAN_TEST_CXXFLAGS_ to XAPIAN_TEST_CXXFLAGS (since the latter is never used) and make it take a variable to add the flags to. Also, fix it to restore CXXFLAGS correctly if the success case contains "break". Wed Aug 06 10:49:46 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/compactor.h: Fix formatting of doxygen comment. Wed Aug 06 10:49:20 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/constants.h: Whitespace tweak. Wed Aug 06 10:08:09 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_version.cc,common/io_utils.h, include/xapian/constants.h: Add DB_FULL_SYNC flag - if this is set for a database, we use the Mac OS X F_FULL_SYNC instead of fdatasync()/fsync()/etc on the version file when committing. Mon Aug 04 05:58:13 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc: Whitespace tweak. Mon Aug 04 05:57:27 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc: When compacting, sync all the tables together at the end. Mon Aug 04 05:42:21 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc,backends/brass/brass_version.cc: Clean up temporary file in BrassVersion::sync() in all failure cases. Mon Aug 04 05:21:02 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/: In DB_DANGEROUS mode, update the version file in-place. Fri Aug 01 02:49:57 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc: Lock the target database while we're compacting to it, which has the side effect of creating the 'flintlock' file there. Fri Aug 01 02:35:35 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/visibility.h: Add link to GCC Visibility wiki page. Thu Jul 31 02:54:40 GMT 2014 Olly Betts <olly@survex.com> * tests/api_weight.cc,weight/bb2weight.cc: Avoid calculating logs of 0 or negative values by clamping the normalise wdf to at most one less than the collection frequency, and the collection size to at least 2. Thu Jul 31 02:48:35 GMT 2014 Olly Betts <olly@survex.com> * tests/api_weight.cc,weight/bb2weight.cc: Fix upper bound for BB2Weight to not be scaled twice by wqf and factor. Wed Jul 30 05:48:46 GMT 2014 Olly Betts <olly@survex.com> * weight/bb2weight.cc: Fix comment typo, and capitalise "Stirling" in comments. Wed Jul 30 05:47:31 GMT 2014 Olly Betts <olly@survex.com> * weight/bb2weight.cc: Whitespace cleanup. Wed Jul 30 05:43:59 GMT 2014 Aarsh Shah <aarshkshah1992@gmail.com> * include/xapian/weight.h,weight/bb2weight.cc: Optimise BB2Weight and fix incorrect upper bound in some cases. Tue Jul 29 07:20:57 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_version.cc: Remove assertion that table sizes are all the same, and redundant assertion that the unserialised blocksize is >= 2048. Fixes set-but-not-used warning in non-assertion build. Mon Jul 28 06:35:45 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_databasereplicator.cc, backends/brass/brass_databasereplicator.h, backends/brass/brass_table.cc: Replace enum table_id with enum Brass::table_type. Mon Jul 28 04:28:12 GMT 2014 Olly Betts <olly@survex.com> * NEWS: Note the freelist implementation fixed #40. Mon Jul 28 04:26:07 GMT 2014 Olly Betts <olly@survex.com> * api/compactor.cc,backends/brass/,backends/dbcheck.cc, tests/api_wrdb.cc: Eliminate the base files, and instead store the root block and freelist pointers in the "iambrass" file. Mon Jul 28 03:22:22 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_databasereplicator.cc: Remove variable which is set but never used. Sun Jul 27 01:56:58 GMT 2014 Olly Betts <olly@survex.com> * backends/dbcheck.cc: Don't try to write output when no output stream is specified in the case where the database open fails and so we can't check changes files. Thu Jul 24 21:12:23 GMT 2014 Olly Betts <olly@survex.com> * NEWS: Update. Fri Jul 11 10:44:48 GMT 2014 Olly Betts <olly@survex.com> * weight/bm25weight.cc: Improve BM25 upper bound in the case when our wdf upper bound > our document length lower bound. Thanks to Craig Macdonald for pointing out this trick. Thu Jul 10 13:29:50 GMT 2014 Olly Betts <olly@survex.com> * weight/bm25weight.cc: Pre-multiply termweight by (param_k1 + 1). Sat Jun 28 11:00:49 GMT 2014 Olly Betts <olly@survex.com> * languages/Makefile.mk: Sort the stop.list files (since loading a sorted list is likely to be more efficient) and remove any duplicate entries. Sat Jun 28 10:48:03 GMT 2014 Olly Betts <olly@survex.com> * languages/stopwords/finnish/stop.txt, languages/stopwords/norwegian/stop.txt: Remove duplicate entries. Sat Jun 28 10:38:56 GMT 2014 Olly Betts <olly@survex.com> * languages/stopwords/russian/stop.txt: Convert from KOI8-R to UTF-8. Mon Jun 23 02:26:24 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: configure.ac: Add hints for which package to install for rst2html Mon Jun 23 00:54:05 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Use AS_MKDIR_P instead of test -d || mkdir. Sun Jun 22 07:32:11 GMT 2014 Olly Betts <olly@survex.com> * NEWS: Update. Sat Jun 21 03:45:19 GMT 2014 Olly Betts <olly@survex.com> * docs/Makefile.am: Use $(MKDIR_P) instead of test -d || mkdir 3 times. Sat Jun 21 03:44:37 GMT 2014 Olly Betts <olly@survex.com> * languages/Makefile.mk: Generate the directories for the minimised stopword files in a VPATH build. Tue Jun 17 13:36:40 GMT 2014 Assem Chelli <assem.ch@gmail.com> * include/xapian/queryparser.h: Add the example of initializing SimpleStopper using a file as a doccomment. Tue Jun 17 13:30:09 GMT 2014 Olly Betts <olly@survex.com> * languages/Makefile.mk: Remove trailing whitespace (some of which breaks the parsing to get the lists of source files to pass to doxygen). Tue Jun 17 13:24:59 GMT 2014 Assem Chelli <assem.ch@gmail.com> * languages/: Add stopwords lists from Snowball plus an Arabic one. Tue Jun 17 11:48:12 GMT 2014 Olly Betts <olly@survex.com> * weight/dlhweight.cc: Fix comment typo. Tue Jun 17 11:46:40 GMT 2014 Aarsh Shah <aarshkshah1992@gmail.com> * weight/dlhweight.cc: Improve the upper bound on DLH ~3-4 times. Tue Jun 17 11:32:22 GMT 2014 Aarsh Shah <aarshkshah1992@gmail.com> * include/xapian/weight.h,weight/dphweight.cc: Combine factor into constants calculated in init(). Improve get_maxpart() a little in the (common) case where the wdf upper bound is less than half the document length upper bound. Tue Jun 17 03:42:36 GMT 2014 Olly Betts <olly@survex.com> * tests/api_unicode.cc,unicode/UnicodeData-README.txt, unicode/UnicodeData.txt: Update to Unicode 7.0.0. Mon Jun 16 14:13:18 GMT 2014 Aarsh Shah <aarshkshah1992@gmail.com> * weight/dphweight.cc: Improved upper bound for DPH. Mon Jun 16 13:58:24 GMT 2014 Aarsh Shah <aarshkshah1992@gmail.com> * include/xapian/weight.h,weight/ifb2weight.cc,weight/ineb2weight.cc, weight/inl2weight.cc: Include 'factor' in the constants we calculate in init_(). Mon Jun 16 13:53:05 GMT 2014 Aarsh Shah <aarshkshah1992@gmail.com> * include/xapian/weight.h,weight/dlhweight.cc: DLH optimization. Mon Jun 16 13:02:24 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Need python3-sphinx to generate docs for python3. Mon Jun 16 12:37:13 GMT 2014 Assem Chelli <assem.ch@gmail.com> * HACKING: Note sphinx-doc now needed (for Python bindings API docs). Mon Jun 16 05:35:03 GMT 2014 Olly Betts <olly@survex.com> * HACKING,configure.ac,docs/Makefile.am,docs/doxygen_api.conf.in: Don't build, ship or install PDF versions of the API docs by default, but provide an easy way for people to build it for themselves if they want it. Mon Jun 16 04:20:40 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Improve the descriptions of the stem_strategy values. (Reported by "oilap" on #xapian) Fri Jun 13 07:52:53 GMT 2014 Olly Betts <olly@survex.com> * weight/dphweight.cc: Fix some comment typos. Fri Jun 13 07:47:32 GMT 2014 Aarsh Shah <aarshkshah1992@gmail.com> * include/xapian/weight.h,weight/dphweight.cc: Optimized DPH code. Fri Jun 13 01:39:07 GMT 2014 Olly Betts <olly@survex.com> * examples/quest.cc: Fix ordering of "bm25" and "bool" in allegedly sorted list, so that --weight bm25 and --weight bool work. Thu Jun 12 13:49:19 GMT 2014 Olly Betts <olly@survex.com> * matcher/extraweightpostlist.h: Add an assertion that get_sumextra() returns a weight contribution which is <= the upper bound which get_maxextra() returns. Thu Jun 12 13:30:55 GMT 2014 Olly Betts <olly@survex.com> * api/leafpostlist.cc,matcher/synonympostlist.cc: Add assertions that get_sumpart() returns a weight contribution which is <= the upper bound which get_maxpart() returns. Wed Jun 11 05:34:16 GMT 2014 Olly Betts <olly@survex.com> * languages/hungarian.sbl: Fix incorrect Unicode codepoints for o-double-acute and u-double-acute in Hungarian Snowball stemmer. Reported by Tom Lane to snowball-discuss. Tue Jun 10 10:59:18 GMT 2014 Aarsh Shah <aarshkshah1992@gmail.com> * include/xapian/weight.h,weight/ineb2weight.cc: Optimized IneB2: Reduced number of calculations in get_sumpart and tightened upper bound. Tue Jun 10 10:23:09 GMT 2014 Olly Betts <olly@survex.com> * weight/ineb2weight.cc: Fix closing brace indentation. Tue Jun 10 09:58:18 GMT 2014 Olly Betts <olly@survex.com> * examples/quest.cc: Rework the weight name lookup to use a table like the other lookups. Change the weight names not to include "weight". Tue Jun 10 08:15:18 GMT 2014 Aarsh Shah <aarshkshah1992@gmail.com> * include/xapian/weight.h,weight/ifb2weight.cc: Optimized upper bound and weighting formula in IfB2Weight. Tue Jun 10 05:34:52 GMT 2014 Aarsh Shah <aarshkshah1992@gmail.com> * include/xapian/weight.h,weight/inl2weight.cc: Optimized upper bound and weighting formula in InL2Weight. Tue Jun 10 05:09:08 GMT 2014 Olly Betts <olly@survex.com> * examples/quest.cc: Add --weight option to allow the weighting scheme to be specified. Thu Jun 05 03:42:51 GMT 2014 Olly Betts <olly@survex.com> * api/omdatabase.cc,tests/api_backend.cc: Fix Database::get_wdf_upper_bound("") not to fail with an assertion. Tue Jun 03 22:59:10 GMT 2014 Olly Betts <olly@survex.com> * matcher/localsubmatch.cc: Use RETURN() instead of return in methods with LOGCALL annotations. Mon Jun 02 13:15:36 GMT 2014 Olly Betts <olly@survex.com> * docs/Makefile.am: Actually ship, process and install geospatial.rst. Mon Jun 02 11:40:27 GMT 2014 Olly Betts <olly@survex.com> * languages/Makefile.mk: Ship collate-sbl script. Thu May 29 12:13:06 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Replace a link to the outdated autotools "goat book" with a link to the "Portable Shell" chapter of the autoconf manual. Thu May 29 03:31:10 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc,api/snipperinternal.h,include/xapian/snipper.h: Allow the user to specify how long a snippet they want. Only add "..." if the sample was actually truncated. Wed May 28 12:52:26 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/snipper.h: Add the standard guards to prevent user code from including this header directly. Tue May 27 04:19:12 GMT 2014 Olly Betts <olly@survex.com> * include/Makefile.mk: Ship include/xapian/snipper.h in tarball. Tue May 27 04:03:09 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc,include/xapian/snipper.h, tests/generate-api_generated: Add missing assignment operator to Xapian::Snipper class. Tue May 27 03:43:03 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/snipper.h: Add missing doxygen comments. Tue May 27 03:41:08 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/snipper.h,tests/api_snipper.cc: Whitespace tweaks. Tue May 27 03:36:19 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc,include/xapian/snipper.h,tests/api_snipper.cc: Remove getters for obtaining info about the relevance model, and instead add a get_description() method which includes all such info in the returned description. Tue May 27 03:06:36 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc: Use an iterator instead of subscripting. Tue May 27 03:03:33 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc,api/snipperinternal.h,include/xapian/snipper.h: Use appropriate named types rather than "int" or "unsigned int". Tue May 27 03:02:48 GMT 2014 Olly Betts <olly@survex.com> * api/snipperinternal.h: Remove constants which are no longer used. Tue May 27 02:48:44 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc,api/snipperinternal.h,include/xapian/snipper.h: Rework Snipper methods set_smoothing_coef() and set_window_size() as optional parameters of generate_snippet(). Tue May 27 02:33:39 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc,api/snipperinternal.h: Just pass rm_docno through to Snipper::Internal::calculate_rm() as a parameter rather than storing it in a member variable. Tue May 27 02:25:41 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc: Use clear() method to empty a container. Tue May 27 02:22:37 GMT 2014 Olly Betts <olly@survex.com> * tests/api_snipper.cc: Drop unused <algorithm>. Tue May 27 02:21:53 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc: Eliminate ret_value variable. Tue May 27 02:20:25 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc: Avoid unnecessary substr operations. Tue May 27 02:17:32 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc: Prefer ++i to i++ in loops. Tue May 27 02:16:14 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc: Use TermIterator::positionlist_count(). Tue May 27 02:14:52 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc,api/snipperinternal.h,include/xapian/snipper.h: Fix typos in comments and variable names in new snipper code. Tue May 27 02:13:09 GMT 2014 Olly Betts <olly@survex.com> * tests/Makefile.am: Hook up new Snipper API tests to actually be used. Tue May 27 02:09:49 GMT 2014 Mihai Bivol <mm.bivol@gmail.com> * tests/api_snipper.cc: Add Snipper API tests. Tue May 27 02:04:26 GMT 2014 Olly Betts <olly@survex.com> * api/snipper.cc: Add missing #include <config.h>. Tue May 27 02:00:32 GMT 2014 Mihai Bivol <mm.bivol@gmail.com> * api/Makefile.mk,api/snipper.cc,api/snipperinternal.h, include/xapian.h,include/xapian/snipper.h: Add Xapian::Snipper class. Mon May 26 05:52:22 GMT 2014 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/brass/brass_postlist.h, backends/chert/chert_postlist.h,backends/database.h: Fix comment typos. Fri May 23 11:59:26 GMT 2014 Olly Betts <olly@survex.com> * api/matchspy.cc: Pass std::string by const reference. Thu May 22 06:31:12 GMT 2014 Olly Betts <olly@survex.com> * matcher/localsubmatch.cc: Fix the check for a term which matches all documents to use the sub-db termfreq, not the combined db termfreq. Thu May 22 06:00:47 GMT 2014 Olly Betts <olly@survex.com> * matcher/localsubmatch.cc: If a term matches all documents and its weight doesn't depend on its wdf, we can optimise it to MatchAll (the previous requirement that maxpart == 0 was unnecessarily strict). Thu May 22 05:52:15 GMT 2014 Olly Betts <olly@survex.com> * matcher/localsubmatch.cc,tests/api_percentages.cc: When we optimise a postlist for a term which matches all documents to use MatchAll, we still need to set a weight object on it to get percentages calculated correctly. Thu May 22 04:08:35 GMT 2014 Olly Betts <olly@survex.com> * api/queryinternal.cc,matcher/localsubmatch.cc, matcher/localsubmatch.h,matcher/queryoptimiser.h: Refactor to group most of the code to create a postlist for a term and set its weight into one place. Mon May 19 00:08:04 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_freelist.cc,tests/api_backend.cc: When we start to write a new freelist block, set the revision on the old one we write out correctly, and don't "-1" out the next block pointer before we write out the block in this case. New regression testcase newfreelistblock1. * backends/brass/brass_freelist.cc: Add assertions that the freelist block has a next pointer when fl_end is a different block to fl. Fri May 16 04:35:21 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_freelist.h: Fix BrassFreeList::swap() to also swap the flw_appending members. Thu May 15 12:51:15 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Add checking config/config.* are up to date to the release checklist. Thu May 15 12:47:28 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Now using automake 1.14.1 to bootstrap snapshots and 1.3.x releases. Wed May 14 03:18:37 GMT 2014 Olly Betts <olly@survex.com> * api/queryinternal.cc: Only serialise parameter where it is actually used. Fri May 09 01:56:48 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Now using doxygen 1.8.7 to bootstrap tarballs. Mon May 05 05:37:32 GMT 2014 Olly Betts <olly@survex.com> * backends/chert/chert_check.cc,backends/chert/chert_dbcheck.cc, backends/dbcheck.cc: Improve DBCHECK_FIX - if fixing a whole database, we now take the revision from the first table we successfully look at, which should be correct in most cases, and is definitely better than trying to determine the revision of each broken table independently. Fri May 02 02:36:05 GMT 2014 Olly Betts <olly@survex.com> * matcher/valuestreamdocument.cc,matcher/valuestreamdocument.h: Put the docid in the sub-database into the ValueStreamDocument object as that's what Xapian::Document::get_docid() is documented to return. (fixes#636, reported by Jeff Rand). Thu May 01 13:25:35 GMT 2014 Olly Betts <olly@survex.com> * api/omdatabase.cc,include/xapian/database.h, matcher/valuestreamdocument.cc: Eliminate Database::get_document_lazily_() completely. Thu May 01 11:48:00 GMT 2014 Olly Betts <olly@survex.com> * backends/database.cc,backends/slowvaluelist.cc, backends/slowvaluelist.h: Don't wrap the subdatabase object in a Xapian::Database for SlowValueList, which mean we can just call open_document(did, true) on it instead of having to use Xapian::Database::get_document_lazily_(). Wed Apr 30 00:31:52 GMT 2014 Olly Betts <olly@survex.com> * api/omdatabase.cc: Explicitly '#include <algorithm>' for std::max(), fixing build with VS2013. (fixes#641) Tue Apr 29 04:28:06 GMT 2014 Olly Betts <olly@survex.com> * common/append_filename_arg.h,net/remoteconnection.h: Comment tweaks. Tue Apr 29 04:26:26 GMT 2014 Olly Betts <olly@survex.com> * tests/harness/unixcmds.cc: Drop code supporting Microsoft Windows 9x which reached EOL in 2006. Thu Apr 17 05:06:21 GMT 2014 Olly Betts <olly@survex.com> * common/autoptr.h: Under C++11, use unique_ptr for AutoPtr. Wed Apr 09 22:16:37 GMT 2014 Olly Betts <olly@survex.com> * api/documentterm.h,api/omdocument.cc: Don't store the termname in OmDocumentTerm - we were only using it in get_description() output and an exception message. Speeds up indexing etext.txt using simpleindex by 0.4%, and should reduce memory usage a bit too. (Change inspired by comments from Vishesh Handa on xapian-devel). Wed Apr 09 05:17:58 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_btreebase.cc,backends/brass/brass_table.cc: Remove comment references to using bitmaps for freespace tracking. Tue Apr 08 04:20:00 GMT 2014 Olly Betts <olly@survex.com> * bin/xapian-delve.cc: Add -A <prefix> option to list all terms with a particular prefix. Mon Apr 07 05:45:06 GMT 2014 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/brass/,backends/chert/, backends/database.cc,backends/database.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/remote/remote-database.cc,backends/remote/remote-database.h, common/remoteprotocol.h,matcher/const_database_wrapper.cc, matcher/const_database_wrapper.h,net/remoteserver.cc, net/remoteserver.h,weight/weightinternal.cc: Fetch termfreq and collection frequency in a single call - shaves ~0.3% off a test search. Thu Apr 03 23:40:45 GMT 2014 Olly Betts <olly@survex.com> * net/serialise.cc,weight/weightinternal.h: Make Weight::Internal track if any max_part values are set, so we don't need to serialise them when they've not been set. Thu Apr 03 05:50:13 GMT 2014 Olly Betts <olly@survex.com> * weight/weightinternal.cc: Enhance Weight::Internal::get_description() to report TermFreqs. Thu Apr 03 05:46:44 GMT 2014 Olly Betts <olly@survex.com> * api/omenquire.cc,api/omenquireinternal.h,common/submatch.h,matcher/, net/remoteserver.cc,net/serialise.cc,weight/weight.cc, weight/weightinternal.cc,weight/weightinternal.h: Eliminate the map<string, TermFreqAndWeight> in MSet::Internal, and instead just preserve the Weight::Internal object there as that holds all the information we need. Thu Apr 03 04:40:09 GMT 2014 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,net/remoteserver.cc, net/serialise.cc,net/serialise.h: Avoid returning Xapian::Weight::Internal by value from unserialise_stats(). Tue Apr 01 22:41:16 GMT 2014 Olly Betts <olly@survex.com> * matcher/localsubmatch.cc,weight/weight.cc,weight/weightinternal.cc, weight/weightinternal.h: Replace Weight::Internal methods to get termfreq, reltermfreq and colltermfreq with a single inlined get_stats() method, which only needs to search the map once. Shaves ~1.6% off the time to do a wildcard query with a lot of terms. Tue Apr 01 22:11:52 GMT 2014 Olly Betts <olly@survex.com> * matcher/localsubmatch.cc: Don't refetch the termfreq if we already have it. Sun Mar 30 22:54:22 GMT 2014 Anish Kanchan <anishgkanchan@gmail.com> * include/xapian/termgenerator.h: Make TermGenerator flags an anonymous enum, and typedef TermGenerator::flags to int for backward compatibility with existing user code which uses it. (fixes #616) Thu Mar 27 05:45:14 GMT 2014 Liu Chi <liuchi09@gmail.com> * include/xapian/stem.h,languages/Makefile.mk, languages/earlyenglish.sbl,languages/stem.cc,tests/api_stem.cc: Add an early english stemmer. Wed Mar 19 02:42:00 GMT 2014 Olly Betts <olly@survex.com> * weight/weightinternal.cc: Fix comment typo. Fri Mar 14 09:53:22 GMT 2014 Olly Betts <olly@survex.com> * HACKING: MacTeX seems to be the best option if using homebrew. Fri Mar 07 23:17:43 GMT 2014 Olly Betts <olly@survex.com> * matcher/maxpostlist.cc: More fixes for --enable.log. Fri Mar 07 23:01:45 GMT 2014 Olly Betts <olly@survex.com> * matcher/maxpostlist.cc: Fix compilation error with --enable-log on. Reported by Anurag Soni. Fri Mar 07 00:39:11 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Fix typo and RST markup. Fri Mar 07 00:21:16 GMT 2014 VcamX <vcamx3@gmail.com> * HACKING: Note packages to install from homebrew on Mac OS X. Fri Mar 07 00:17:01 GMT 2014 VcamX <vcamx3@gmail.com> * backends/dbcheck.cc,common/pack.h: Fix warnings with clang 5.0. Tue Mar 04 04:25:11 GMT 2014 Olly Betts <olly@survex.com> * api/queryinternal.cc,api/queryinternal.h: Only simplify OP_SYNONYM with a single subquery if that subquery is a term or MatchAll. Tue Mar 04 04:12:58 GMT 2014 Olly Betts <olly@survex.com> * tests/api_none.cc: Test OPs with a single MatchAll subquery. Mon Mar 03 22:59:55 GMT 2014 Mayank Chaudhary <mayankchaudhary.iitr@gmail.com> * docs/Makefile.am,docs/bm25.rst,docs/intro_ir.rst: Convert equations in rst docs to use LaTeX via the math role and directive. Fri Feb 28 06:12:36 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.h, backends/chert/chert_positionlist.h: Remove next_internal() methods which were declared but not defined or used (left-overs from flint). Reported by "Hurricane Tong". Wed Feb 26 01:33:50 GMT 2014 Olly Betts <olly@survex.com> * api/query.cc,api/queryinternal.cc,api/queryinternal.h, examples/quest.cc,include/xapian/query.h,matcher/Makefile.mk, matcher/maxpostlist.cc,matcher/maxpostlist.h, queryparser/queryparser.cc,tests/api_none.cc,tests/api_opsynonym.cc: Implement new OP_MAX query operator, which returns the maximum weight of any of its subqueries. (see #360) Wed Feb 26 00:48:23 GMT 2014 Olly Betts <olly@survex.com> * matcher/localsubmatch.cc,matcher/synonympostlist.cc, matcher/synonympostlist.h: In SynonymPostList, avoid fetching the doclength if the weighting scheme doesn't need it and the calculated wdf for the synonym in <= doclength_lower_bound for the current subdatabase. (fixes #360) Wed Feb 26 00:12:22 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_changes.cc: Fix compilation without --enable-assertions. Tue Feb 25 02:33:14 GMT 2014 Olly Betts <olly@survex.com> * examples/quest.cc: Add --default-op option. Tue Feb 25 01:56:34 GMT 2014 Olly Betts <olly@survex.com> * examples/quest.cc: Whitespace tweak. Mon Feb 24 06:33:37 GMT 2014 Olly Betts <olly@survex.com> * tests/api_opsynonym.cc: Refactor symonym1 so that the test data is specified by an array, rather than built up by code, so that the actual testcases can be more easily seen. Sun Feb 23 01:22:56 GMT 2014 Olly Betts <olly@survex.com> * common/bitstream.cc,common/bitstream.h: Add comments to better document BitReader and BitWriter. (fixes#252) Sun Feb 23 00:35:17 GMT 2014 Olly Betts <olly@survex.com> * HACKING: Stop trying to explicitly list compilers which support __builtin_expect (LLVM does too, others may as well). Fri Feb 21 01:34:59 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/,backends/dbcheck.cc,common/, docs/replication_protocol.rst,tests/api_backend.cc, tests/api_replicate.cc: Keep track of unused blocks in the Btrees using freelists rather than bitmaps. Fri Feb 21 01:21:45 GMT 2014 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Expand crashrecovery1 to check that the expected base files exist and ones which shouldn't exist don't. Thu Feb 20 00:34:47 GMT 2014 Olly Betts <olly@survex.com> * HACKING: makeindex is now in Debian package texlive-binaries. Wed Feb 19 06:50:58 GMT 2014 Olly Betts <olly@survex.com> * docs/sorting.rst: Fix incorrect parameter types in Xapian::Weight subclass example. Tue Feb 18 23:34:19 GMT 2014 Olly Betts <olly@survex.com> * api/replication.cc: Fix conditional for enabling replication code - if chert is disabled but brass isn't, we should still enable it. Tue Feb 18 01:33:28 GMT 2014 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Move comment to in front of the function it is actually talking about. Sun Feb 16 04:53:57 GMT 2014 Olly Betts <olly@survex.com> * xapian-config.in: Add some missing shell quoting identified by shellcheck. Sun Feb 16 00:53:01 GMT 2014 Olly Betts <olly@survex.com> * docs/quickstart.rst: Split --cxxflags and --libs for portability (noted by "Hurricane Tong" on xapian-devel). Sat Feb 15 04:59:37 GMT 2014 Olly Betts <olly@survex.com> * docs/glossary.rst: Add definition of "collection frequency". Fri Feb 14 22:24:26 GMT 2014 Olly Betts <olly@survex.com> * NEWS: Update. Fri Jan 17 03:50:44 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_databasereplicator.cc, backends/brass/brass_table.cc, backends/chert/chert_databasereplicator.cc, backends/chert/chert_table.cc,common/io_utils.cc,common/io_utils.h: Add io_read_block() and io_write_block() as wrappers around pread() and pwrite() (or lseek() plus read() and lseek() plus write()). Sat Jan 11 06:32:19 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Sun C++ seems to explicitly need Crun now. Sat Jan 11 02:23:06 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Fix checks for Sun C++. Fri Jan 10 22:38:01 GMT 2014 Olly Betts <olly@survex.com> * matcher/multimatch.cc: timer_create() doesn't handle CLOCK_MONOTONIC on Solaris, so use CLOCK_REALTIME there instead. Fri Jan 10 11:22:37 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Improve handling of Sun's C++ compiler - trick libtool into not adding -library=Cstd, and prefer -library=stdcxx4 if supported. Fri Jan 10 11:16:15 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Add XAPIAN_TEST_CXXFLAGS macro, and use it to check the flags to cxx actually work. Fri Jan 10 10:56:17 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Apply missing functional part of previous change. Fri Jan 10 10:46:53 GMT 2014 Olly Betts <olly@survex.com> * configure.ac: Avoid using grep -e as /usr/bin/grep on Solaris doesn't support it. Fri Jan 10 03:30:10 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_databasereplicator.cc: Whitespace tweak. Thu Jan 09 11:06:05 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_cursor.h,backends/brass/brass_database.cc, backends/chert/chert_database.cc, backends/chert/chert_databasereplicator.cc, backends/inmemory/inmemory_database.h,common/log2.h, common/realtime.h,matcher/multimatch.cc,weight/pl2weight.cc, weight/tradweight.cc: Fix compilations issues with Sun's C++ compiler (mostly missing library headers). Thu Jan 09 06:00:03 GMT 2014 Olly Betts <olly@survex.com> * backends/dbcheck.cc: Skip absent lazily created tables in the case where out == NULL too. Thu Jan 09 05:31:40 GMT 2014 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Use Xapian::Database::check() in cursordelbug1. (partly addresses #238) Thu Jan 09 05:30:45 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/database.h: Make second and third parameters to Xapian::Database::check() optional, defaulting to a quiet check. Thu Jan 09 02:52:54 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/,backends/chert/,backends/dbcheck.cc, bin/xapian-check.cc,include/xapian/database.h: Alter Database::check() to take its "out" parameter as a pointer to std::ostream instead of a reference, and make passing NULL means "do not produce output". Remove the form without the "out" parameter, which used to mean "send output to stdout" - if you want this, just pass "&std::cout" in. Thu Jan 09 01:50:14 GMT 2014 Olly Betts <olly@survex.com> * include/xapian/database.h: OS X SDK check() macro takes one argument, so there's no need to protect check() methods with more than one argument. * include/xapian/valueiterator.h: Protect the one check() method in our API which does take one argument. Wed Jan 08 11:48:58 GMT 2014 Olly Betts <olly@survex.com> * api/smallvector.cc,api/smallvector.h: Include cstddef rather than cstring for std::size_t, as it's a simpler header. Don't assume size_t will be available unqualified. Add SmallVector::size_type typedef. Mon Jan 06 11:54:53 GMT 2014 Olly Betts <olly@survex.com> * net/progclient.cc: Fix xlC warning. Mon Jan 06 00:31:19 GMT 2014 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/chert/chert_table.cc: Validate dir_end when reading a block. (fixes #592) Wed Jan 01 01:20:46 GMT 2014 Olly Betts <olly@survex.com> * docs/Makefile.am: Fix handling of MAINTAINER_NO_DOCS. Mon Dec 30 05:33:59 GMT 2013 Olly Betts <olly@survex.com> * tests/api_unicode.cc: Add regression test for incorrect results from Unicode::tolower() for 4 code points (U+01C5, U+01C8, U+01CB, U+01F2) with Xapian 1.2.16 and earlier. Sun Dec 29 05:46:14 GMT 2013 Olly Betts <olly@survex.com> * docs/postingsource.rst: Clarify a couple of points (reported by "vHanda" on IRC). Tue Dec 24 02:00:26 GMT 2013 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/perftest/Makefile.mk: Actually use @NO_INSTALL@. Tue Dec 24 01:59:05 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Use -no-fast-install on platforms where -no-install causes libtool to emit a warning. Mon Dec 23 04:17:35 GMT 2013 Olly Betts <olly@survex.com> * HACKING: Drop note about needing git-svn if you're using git - bootstrap now only uses git-svn if your Xapian tree was checked out using git-svn. Mon Dec 23 04:16:58 GMT 2013 Olly Betts <olly@survex.com> * HACKING: Note that Tcl is now needed to generate unicode/unicode-data.cc. Mon Dec 23 04:01:05 GMT 2013 Olly Betts <olly@survex.com> * tests/api_unicode.cc: Add a test that a character due to be added in Unicode 7.0.0 is categorised as "unassigned". Mon Dec 23 03:50:37 GMT 2013 Olly Betts <olly@survex.com> * unicode/Makefile.mk,unicode/uniParse.tcl: Adjust to not need a symlink. Give the generated file a new name, since the "tcl" prefix doesn't make much sense here, and it avoids issues with transitioning from a file in version control to one which is generated. Mon Dec 23 03:32:40 GMT 2013 Olly Betts <olly@survex.com> * unicode/: Add the source data file and scripts to generate tclUniData.cc. Mon Dec 23 02:20:57 GMT 2013 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Mon Dec 23 01:04:26 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Note that 1.2.17 is required for the correct sense of the reverse parameter to Enquire methods set_sort_by_relevance_then_value() and set_sort_by_relevance_then_key(). Mon Dec 23 00:30:53 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_btreebase.cc: Fix comment documenting what's in the base file (the sequential flag was missing). Fri Dec 20 03:47:56 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_termlisttable.h,include/xapian/constants.h, tests/api_backend.cc: Add Xapian::DB_NO_TERMLIST flag to allow creating a database without a termlist (see #181). Fri Dec 20 03:04:51 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/unicode.h,unicode/tclUniData.cc: Update Unicode character database to 6.3.0. * tests/api_unicode.cc: Add tests for U+061C, added in 6.3.0. Fri Dec 20 00:20:01 GMT 2013 Olly Betts <olly@survex.com> * backends/dbfactory.cc: Clear any backend flag passed in before passing the flags when opening stub databases, fixing test failure in replicate1 caused by recent change adding Xapian::DB_BACKEND_STUB. Thu Dec 19 23:11:32 GMT 2013 Olly Betts <olly@survex.com> * tests/api_backend.cc: Simplify database creation in testcase blocksize1. Thu Dec 19 23:10:37 GMT 2013 Olly Betts <olly@survex.com> * backends/dbfactory.cc: Tweak conditionals around $XAPIAN_PREFER_BRASS check. Thu Dec 19 23:09:30 GMT 2013 Olly Betts <olly@survex.com> * api/replication.cc,backends/dbfactory.cc,docs/deprecation.rst, docs/overview.rst,include/xapian/constants.h, include/xapian/dbfactory.h,tests/api_backend.cc,tests/api_db.cc: Add Xapian::DB_BACKEND_STUB and deprecate Xapian::Auto::open_stub(). Thu Dec 19 06:46:15 GMT 2013 Olly Betts <olly@survex.com> * docs/deprecation.rst: Update example version to a more recent one. Thu Dec 19 04:01:31 GMT 2013 Olly Betts <olly@survex.com> * docs/deprecation.rst,include/xapian/dbfactory.h,tests/api_db.cc, tests/harness/: Deprecate Xapian::Brass::open() and Xapian::Chert::open() in favour of the new Xapian::DB_BACKEND_BRASS and Xapian::DB_BACKEND_CHERT flags. Thu Dec 19 01:17:02 GMT 2013 Olly Betts <olly@survex.com> * api/compactor.cc,backends/brass/,backends/chert/chert_database.cc, backends/chert/chert_database.h,backends/database.h, backends/dbcheck.cc,backends/dbfactory.cc,include/xapian/constants.h, include/xapian/database.h,include/xapian/dbfactory.h: Add flags which can be bitwise OR-ed into the second argument when opening a writable datase: DB_NO_SYNC (to disable use of fsync, etc) and DB_DANGEROUS (to enable in-place updates). WritableDatabase can now be constructed from just a filename (defaulting to opening the database with DB_CREATE_OR_OPEN). Thu Dec 19 00:11:26 GMT 2013 Olly Betts <olly@survex.com> * tests/harness/backendmanager_multi.cc: Fix comment typo. Wed Dec 18 23:52:49 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/chert/chert_table.cc, tests/api_backend.cc: Fix handling of invalid block sizes - the size gets fixed as documented, but the uncorrected size was passed to the base file. Also, abort() was called on 0. Wed Dec 18 22:48:18 GMT 2013 Olly Betts <olly@survex.com> * api/replication.cc,backends/brass/brass_check.cc, backends/brass/brass_database.cc,backends/chert/chert_check.cc, backends/chert/chert_database.cc,backends/database.h, backends/dbcheck.cc,bin/xapian-tcpsrv.cc,include/Makefile.mk, include/xapian.h,include/xapian/database.h,net/remoteserver.cc: Move Xapian::DB_* and Xapian:DBCHECK_* constants into xapian/constants.h. Wed Dec 18 22:30:23 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc,backends/chert/chert_database.cc: Include xapian/*.h with "" not <> inside the library. Wed Dec 18 22:17:15 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc,backends/chert/chert_compact.cc: Include only the needed xapian/*.h headers (rather than xapian.h) to reduce inter-file dependencies within the library. Wed Dec 18 03:24:22 GMT 2013 Olly Betts <olly@survex.com> * tests/runtest.in: Report if a test is being run under eatmydata, like we do for running under valgrind. Wed Dec 18 02:43:47 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_version.h,backends/chert/chert_version.h: Comment out unused set_uuid() and set_uuid_string() methods - they may be useful in future, but we aren't currently using them. Also make set_uuid() take a const pointer. Wed Dec 18 02:40:18 GMT 2013 Olly Betts <olly@survex.com> * HACKING,configure.ac,tests/runtest.in: Automatically probe for and hook in eatmydata to the testsuite using the wrapper script it now includes. Wed Dec 18 01:21:07 GMT 2013 Olly Betts <olly@survex.com> * HACKING: Update website update info. Sat Dec 14 06:01:33 GMT 2013 Olly Betts <olly@survex.com> * weight/dlhweight.cc,weight/dphweight.cc: Whitespace tweak. Fri Dec 13 21:06:28 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_cursor.cc,backends/brass/brass_cursor.h, backends/brass/brass_table.cc: Reference count blocks in the btree cursor, so cursors can cheaply share blocks. This can significantly reduce the amount of memory used for cursors for queries which contain a lot of terms (e.g. wildcards which expand to a lot of terms). Thu Dec 12 22:37:22 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_btreebase.cc: Add assertions that we don't read beyond the end of the freespace bitmap. Thu Dec 12 22:25:04 GMT 2013 Olly Betts <olly@survex.com> * matcher/msetcmp.cc: Fix sense of reverse parameter to Enquire::set_sort_by_relevance_then_value. Reported by "boomboo" on IRC. * tests/api_db.cc: Fix testcase sortrel1. Thu Dec 12 19:47:46 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/: Make Brass::Cursor members n and p private. Thu Dec 12 19:46:23 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_check.cc,backends/brass/brass_table.cc: Add const to a few pointers we only read through. Thu Dec 12 09:04:13 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Improve reporting of GCC version. Thu Dec 12 09:01:26 GMT 2013 Olly Betts <olly@survex.com> * tests/api_query.cc,unicode/description_append.cc: Escape \x7f in object descriptions. Wed Dec 11 23:40:37 GMT 2013 Olly Betts <olly@survex.com> * api/leafpostlist.cc,api/leafpostlist.h,backends/brass/, matcher/localsubmatch.cc,matcher/localsubmatch.h, matcher/queryoptimiser.h: Under brass, optimise the turning of a query into a postlist to copy the cursor blocks for the previous term's postlist. This is particularly effective for wildcard queries which expand to a lot of terms. Wed Dec 11 22:05:34 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_postlist.h: We don't have a BrassModifiedPostList class, so we don't need protected members in BrassPostList for it to use. Wed Dec 11 04:46:58 GMT 2013 Olly Betts <olly@survex.com> * weight/bm25weight.cc: Fix case where (k1 == 0 || b == 0) but k2 != 0. Reported by "boomboo" on IRC. * tests/api_weight.cc: Add regression test bm25weight5. Wed Dec 04 05:44:29 GMT 2013 Olly Betts <olly@survex.com> * NEWS: Update from 1.2.16 and ChangeLog. Mon Dec 02 02:34:43 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Clarify version of MSVCRT which we require for mingw. Fri Nov 29 04:21:46 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Fix typo in previous commit. Fri Nov 29 03:48:24 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Report GCC version. Wed Nov 27 04:06:32 GMT 2013 Olly Betts <olly@survex.com> * Backport change from brass: * backends/chert/chert_table.cc: Check if the block we want is in the internal cursor for a read-only table too - it isn't needed for correctness in this case, but it is more efficient not to reload a block which we already have. Sun Nov 24 23:05:37 GMT 2013 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Fri Nov 22 20:18:43 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_databasereplicator.cc, backends/brass/brass_databasereplicator.h: compression_stream.h isn't used in brass_databasereplicator.h, so move include to brass_databasereplicator.cc. Fri Nov 22 01:22:32 GMT 2013 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Minor tweak to structuring of tests in check_infix(). Sat Nov 09 08:51:27 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc: Check if the block we want is in the internal cursor for a read-only table too - it isn't needed for correctness in this case, but it is more efficient not to reload a block which we already have. Sat Nov 09 08:48:02 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/: Use a separate cursor for each position list - now we're ordering the position B-tree by term first, phrase matching was causing the cursor to cycle between disparate areas of the B-tree, and it would have to reread the same blocks over and over. Fri Nov 08 09:57:52 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/chert/chert_table.cc: Add check for the database being closed in read_block(), which fixes pythontest.py testcase valuemods. Fri Nov 08 08:58:48 GMT 2013 Olly Betts <olly@survex.com> * api/maptermlist.h,api/termlist.cc,api/termlist.h,backends/brass/: Avoid two vector copies when storing term positions in most common cases. Fri Nov 08 08:57:07 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Bump LIBRARY_VERSION_INFO. Thu Nov 07 16:14:23 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc, backends/brass/brass_positionlist.cc, backends/brass/brass_positionlist.h: There's no need to pass around an end PositionIterator, and it's actually slightly more efficient not to. Thu Nov 07 16:11:27 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/: Batch up positionlist changes in memory and write them out in key order. Thu Nov 07 15:30:48 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.h,backends/brass/brass_version.cc: Change positionlist keys to be ordered by term first rather than docid first, which helps phrase searching significantly. Wed Nov 06 17:12:29 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_values.cc,backends/brass/brass_values.h, backends/chert/chert_values.cc,backends/chert/chert_values.h: Reuse a cursor for reading values from valuestreams rather than creating a new one each time. This can dramatically reduce the number of blocks redundantly reread when sorting by value. The rereads will generally get served from VM cache, but there's still an overhead to that. Tue Nov 05 16:28:08 GMT 2013 Olly Betts <olly@survex.com> * api/query.cc,api/queryinternal.cc,api/queryinternal.h, include/xapian/query.h,tests/api_query.cc: Having a separate 'type' enum seems to just make for clumsier code, so merged the LEAF_* values into the 'op' enum, add LEAF_MATCH_ALL and LEAF_MATCH_NOTHING and return when appropriate, and add test coverage. Tue Nov 05 15:20:15 GMT 2013 Olly Betts <olly@survex.com> * api/query.cc,api/queryinternal.cc,api/queryinternal.h, include/xapian/query.h: Add methods to allow introspection on Query objects - currently you can read the type/operator, how many subqueries there are, and get a particular subquery. For a query which is a term, Query::get_terms_begin() allows you to get read the term. (see #159) Mon Nov 04 19:06:38 GMT 2013 Olly Betts <olly@survex.com> * examples/quest.cc: Add --check-at-least option to quest. Thu Oct 24 04:38:52 GMT 2013 Olly Betts <olly@survex.com> * configure.ac,tests/soaktest/soaktest.cc: If srandom() and random() aren't available, use srand() and rand() instead. Thu Oct 24 02:34:06 GMT 2013 Olly Betts <olly@survex.com> * common/safewindows.h: Define WINVER to 0x0500 if not otherwise set, so we get new interfaces like MEMORYSTATUSEX under mingw. Thu Oct 24 02:11:36 GMT 2013 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc: Need cstdlib for free under mingw. Thu Oct 24 02:11:05 GMT 2013 Olly Betts <olly@survex.com> * configure.ac,tests/api_replicate.cc,tests/queryparsertest.cc, tests/termgentest.cc: Probe for _putenv_s() in configure as mingw doesn't provide it. Thu Oct 24 01:23:14 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Don't unset CC as it's used by libtool under mingw. Thu Oct 24 01:17:07 GMT 2013 Olly Betts <olly@survex.com> * net/tcpserver.cc: Fix mingw build error. Thu Oct 24 01:00:18 GMT 2013 Olly Betts <olly@survex.com> * common/omassert.h: Suppress "unused local typedef" warnings from GCC 4.8. Wed Oct 23 20:00:05 GMT 2013 Olly Betts <olly@survex.com> * common/realtime.h: Rework RealTime::to_timeval() under __WIN32__ to avoid having to pull in safewinsock2.h, Wed Oct 23 12:39:53 GMT 2013 Olly Betts <olly@survex.com> * common/realtime.h: Under __WIN32__, we need safewinsock2.h for struct timeval. Wed Oct 23 10:21:02 GMT 2013 Olly Betts <olly@survex.com> * common/omassert.h: If the compiler supports C++11, use static_assert to implement CompileTimeAssert. Wed Oct 23 10:02:50 GMT 2013 Olly Betts <olly@survex.com> * common/realtime.h,net/tcpclient.cc: Fix mingw build errors. Wed Oct 23 09:50:00 GMT 2013 Olly Betts <olly@survex.com> * common/win32_uuid.cc: Fix mingw compiler warning. Wed Oct 23 04:19:05 GMT 2013 jiangwen jiang <jiangwen127@gmail.com> * api/omdocument.cc,api/query.cc: Add more LOGCALL annotations. Tue Oct 22 03:09:48 GMT 2013 Olly Betts <olly@survex.com> * HACKING: Note we now use doxygen 1.8.5 for 1.3.x snapshots and releases. Mon Oct 21 01:55:48 GMT 2013 Olly Betts <olly@survex.com> * queryparser/lemon.c: Sync change from upstream: "Fix a harmless compiler warning in lemon.c." Generated code is identical, at least on x86-64 Debian wheezy. Mon Oct 21 01:52:12 GMT 2013 Olly Betts <olly@survex.com> * queryparser/lemon.c: Sync change from upstream: "In the lemon parser generator, change all hashes to unsigned to avoid potential problems with signed integer overflow." Generated code is identical, at least on x86-64 Debian wheezy. Sat Oct 19 06:38:48 GMT 2013 Olly Betts <olly@survex.com> * backends/flint_lock.cc: Under cygwin, use cygwin_conv_path() if using a new enough cygwin version, rather than the now deprecated cygwin_conv_to_win32_path(). Reported by "Haroogan" on the xapian-devel mailing list. Sat Oct 19 00:36:06 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Simplify test for uuid_unparse_lower(). (see #626) Sat Oct 19 00:21:37 GMT 2013 Olly Betts <olly@survex.com> * Makefile.am,configure.ac: Split XAPIAN_LIBS out of XAPIAN_LDFLAGS. (fixes#626) Fri Oct 18 11:51:40 GMT 2013 Olly Betts <olly@survex.com> * geospatial/geoencode.cc: Use lround() instead of round(), since we want the result as an int. GCC 4.4.3 seems to optimise to use lround() anyway, but other compilers may not. Fri Oct 18 11:44:10 GMT 2013 Olly Betts <olly@survex.com> * geospatial/geoencode.cc: Include <math.h> for round() (fixes #628). Tue Oct 15 03:16:04 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/enquire.h,include/xapian/unicode.h: Mark internal methods which are public for implementation convenience as "@private" for doxygen where they weren't already. Mon Oct 14 09:45:13 GMT 2013 Olly Betts <olly@survex.com> * include/xapian.h: Mark vinfo struct and get_vinfo_() function as @private for doxygen. Thu Oct 10 10:36:45 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/: Methods and functions which take a string to unserialise now consistently call that parameter "serialised". Mon Oct 07 04:08:14 GMT 2013 Olly Betts <olly@survex.com> * tests/zlib-vg.c: Fix two warnings when compiled with clang. Fri Oct 04 19:46:16 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_check.cc,backends/chert/chert_check.cc: Correct failure message which talks above the root block when it's actually testing a leaf key. Fri Oct 04 04:51:39 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_check.cc,backends/chert/chert_check.cc: Escape invalid UTF-8 data in keys and tags reported by xapian-check, using the same code we use to clean up string in object descriptions. Fri Oct 04 04:50:52 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_check.cc,backends/chert/chert_check.cc: Fix xapian-check to report block numbers correctly for links within the B-tree. Thu Oct 03 07:35:44 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_dbcheck.cc,backends/chert/chert_dbcheck.cc: If the METAINFO key is missing, only report it once per table. Sat Sep 28 09:27:51 GMT 2013 Olly Betts <olly@survex.com> * include/xapian.h: Hide struct vinfo and get_vinfo_() from SWIG and doxygen. Sat Sep 28 06:25:44 GMT 2013 Olly Betts <olly@survex.com> * api/version.cc,include/xapian.h: Reimplement version functions to use a single function in libxapian which returns a pointer to a static const struct containing the version information, with inline wrappers in the API header which call this. This means we only need one relocation instead of 4. * tests/api_none.cc: New testcase version1 to provide coverage for the version API functions. Fri Sep 27 22:53:15 GMT 2013 Olly Betts <olly@survex.com> * api/omenquire.cc,include/xapian/enquire.h,tests/api_anydb.cc: Fix backwards compatibility get_eset() wrapper to allow for the optional parameters which were supported before. Also, just inline it from the API header. Fri Sep 27 11:19:50 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * api/omenquire.cc,api/omenquireinternal.h,expand/, include/xapian/enquire.h,tests/api_anydb.cc: Add support for the Bo1 query expansion scheme. Thu Sep 26 04:11:49 GMT 2013 Olly Betts <olly@survex.com> * tests/api_weight.cc: Don't pass integer arguments to log() to avoid ambiguity errors with some compilers. (fixes #627) Thu Sep 26 02:02:24 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/,backends/chert/,backends/dbcheck.cc: Fix database consistency checking to always open all the tables at the same revision, which could lead to false errors being reported after a commit interrupted by the process being killed or the machine crashing. Reported by Joey Hess in https://bugs.debian.org/724610 Wed Sep 25 02:21:29 GMT 2013 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add testcase for FieldProcessor on boolean prefix with quoted contents. Wed Sep 25 02:16:55 GMT 2013 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix comment typo. Wed Sep 25 02:13:20 GMT 2013 Olly Betts <olly@survex.com> * api/: Split QueryVector into a template SmallVector class and then 'typedef SmallVector<Query> QueryVector;', as it would be good to use the same approach for Database internals. Tue Sep 24 09:59:05 GMT 2013 Olly Betts <olly@survex.com> * .gitignore,xapian-core.spec.in: Update two references to .tar.gz files. Tue Sep 24 09:54:42 GMT 2013 Olly Betts <olly@survex.com> * docs/install.rst: Fix typo (bogus extra 'and'); Tue Sep 24 09:54:01 GMT 2013 Olly Betts <olly@survex.com> * configure.ac,docs/install.rst: Compress source tarballs with xz instead of gzip. Mon Sep 23 12:58:23 GMT 2013 Olly Betts <olly@survex.com> * common/realtime.h,configure.ac,matcher/multimatch.cc: Implement RealTime::sleep() using nanosleep() where it's available, since it has a simpler API and a finer granularity than select(). Mon Sep 23 12:27:37 GMT 2013 Olly Betts <olly@survex.com> * common/realtime.h,net/remoteconnection.cc,net/tcpclient.cc: Add RealTime::to_timeval() helper function to break a time interval as a double into a struct timeval. This uses modf() which gives more compact and faster code than using fmod() as we previously did. Mon Sep 23 08:20:03 GMT 2013 Olly Betts <olly@survex.com> * common/realtime.h,configure.ac: Use clock_gettime() to implement RealTime::now() where it's available. Mon Sep 16 11:53:28 GMT 2013 Olly Betts <olly@survex.com> * api/,backends/brass/brass_postlist.cc, backends/remote/net_postlist.cc,backends/slowvaluelist.cc, expand/esetinternal.cc,include/xapian/unicode.h, matcher/valuegepostlist.cc,matcher/valuerangepostlist.cc, tests/api_query.cc,tests/queryparsertest.cc,unicode/: Fix all get_description() methods to always return UTF-8 text. (fixes #620) Mon Sep 16 11:47:16 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Document how Utf8Iterator handles invalid UTF-8. Mon Sep 16 11:19:48 GMT 2013 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add test coverage for explicit synonym of a term with a prefix (e.g. ~foo:search). Fri Sep 06 07:11:29 GMT 2013 Olly Betts <olly@survex.com> * api/valueiterator.cc,tests/api_backend.cc: Don't segfault is skip_to() or check() are called on a ValueIterator which is already at_end(). Fri Sep 06 06:52:49 GMT 2013 Olly Betts <olly@survex.com> * api/positioniterator.cc,api/postingiterator.cc,api/termiterator.cc, tests/api_backend.cc: Don't segfault if skip_to() is called on an iterator which is already at_end(). Reported by David Bremner. Wed Sep 04 04:52:50 GMT 2013 Olly Betts <olly@survex.com> * weight/weightinternal.cc: TermFreqs now tracks collection frequency, so make TermFreqs::get_description() report it. Thu Aug 29 01:30:45 GMT 2013 Olly Betts <olly@survex.com> * matcher/andmaybepostlist.cc,matcher/multiandpostlist.h: Fix assertion failure for when an OrPostList decays to an AndPostList - the ordering of the subqueries by estimated termfreq may not be the same as it was when the OrPostList was constructed, as the subqueries may themselves have decayed. Reported by Michel Pelletier. Fri Aug 23 03:17:33 GMT 2013 Olly Betts <olly@survex.com> * m4-macros/xapian-1.3.m4: Handle empty or unset XAPIAN_CONFIG correctly. Mon Jul 15 12:01:17 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * api/registry.cc,include/xapian/weight.h,tests/api_nodb.cc, tests/api_weight.cc,weight/Makefile.mk,weight/dphweight.cc: Add an implementation of DfR weighting scheme DPH. Sun Jul 14 07:39:37 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/weight.h,weight/pl2weight.cc: Optimise PL2Weight to only calculate 2 logs per call to get_sumpart() instead of 4 by precalculating three document-independent values in init(). Sun Jul 14 07:04:52 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * api/registry.cc,include/xapian/weight.h,tests/api_nodb.cc, tests/api_weight.cc,weight/Makefile.mk,weight/pl2weight.cc: Add an implementation of DfR weighting scheme PL2. Sat Jul 13 11:24:20 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * weight/tfidfweight.cc: Explicitly cast argument of log() to double, as xlC doesn't manage to chose an overloaded form to call otherwise. Sat Jul 13 07:29:25 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * include/xapian/weight.h,tests/api_weight.cc, tests/generate-api_generated,weight/: Fix new weighting schemes to work with OP_SCALE_WEIGHT. Wed Jul 10 12:57:01 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * api/registry.cc,include/xapian/weight.h,tests/api_nodb.cc, tests/api_weight.cc,weight/Makefile.mk,weight/dlhweight.cc: Add an implementation of DfR weighting scheme DLH. Wed Jul 10 07:38:15 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * expand/expandweight.h: Track collection_freq and totlen restricted to the RSet in preparation for supporting DfR query expansion schemes. Mon Jul 08 06:27:05 GMT 2013 Olly Betts <olly@survex.com> * weight/Makefile.mk: Normalise whitespace in BB2 change. Sun Jul 07 12:29:48 GMT 2013 Olly Betts <olly@survex.com> * weight/bb2weight.cc: New file for BB2Weight which I failed to commit. Sun Jul 07 11:51:45 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * api/registry.cc,include/xapian/weight.h,tests/api_nodb.cc, tests/api_weight.cc,weight/Makefile.mk: Add implementation of DfR weighting scheme BB2. Sat Jul 06 02:07:13 GMT 2013 Olly Betts <olly@survex.com> * api/omdatabase.cc: Fix add_document() to always use exactly the next docid after get_lastdocid() in the multidatabase case. Fri Jul 05 06:19:55 GMT 2013 Olly Betts <olly@survex.com> * backends/dbfactory.cc: Allow open a stub database containing multiple subdatabases as a WritableDatabase. Thu Jul 04 06:29:56 GMT 2013 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc,tests/harness/index_utils.cc, tests/harness/index_utils.h: Simplify the FileIndexer class. Thu Jul 04 06:09:06 GMT 2013 Olly Betts <olly@survex.com> * tests/harness/backendmanager_multi.cc,tests/harness/index_utils.cc, tests/harness/index_utils.h: Make use of the new writable multidatabase feature to simplify the multi-database handling in the test harness. Wed Jul 03 13:58:46 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * api/registry.cc,include/xapian/weight.h,tests/api_nodb.cc, tests/api_weight.cc,weight/Makefile.mk,weight/ineb2weight.cc: Add implementation of DfR weighting scheme IneB2. Wed Jul 03 13:39:11 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/weight.h: "please refer:" -> "please refer to:" Wed Jul 03 13:31:59 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * api/registry.cc,include/xapian/weight.h,tests/api_nodb.cc, tests/api_weight.cc,weight/Makefile.mk,weight/ifb2weight.cc: Add implementation of DfR weighting scheme IfB2. Wed Jul 03 12:42:23 GMT 2013 Olly Betts <olly@survex.com> * common/safeuuid.h: Make a few tweaks to support AIX. Wed Jul 03 12:33:40 GMT 2013 Olly Betts <olly@survex.com> * common/safeuuid.h: Remove bogus take-address-of from uuid handling code for netbsd and openbsd. Wed Jul 03 11:56:05 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * api/registry.cc,include/xapian/weight.h,tests/api_nodb.cc, tests/api_weight.cc,weight/Makefile.mk,weight/inl2weight.cc: Add implementation of DfR weighting scheme InL2. Wed Jul 03 08:20:26 GMT 2013 Olly Betts <olly@survex.com> * api/error.cc,common/Makefile.mk,common/safenetdb.h,net/tcpclient.cc, net/tcpserver.cc: AIX needs _USE_IRS defined for hstrerror, so create safenetdb.h to take care of that. Wed Jul 03 07:34:13 GMT 2013 Olly Betts <olly@survex.com> * common/safeuuid.h: Add missing '#include <cstdlib>' and qualify free with std. Tue Jul 02 23:44:48 GMT 2013 Olly Betts <olly@survex.com> * api/omdatabase.cc: Add support for WritableDatabase with multiple subdatabases. Tue Jul 02 23:40:08 GMT 2013 Olly Betts <olly@survex.com> * api/omdatabase.cc,tests/api_none.cc: Trying to perform operations on a database with no subdatabases now throws InvalidOperationError not DocNotFoundError. Tue Jul 02 13:22:31 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * tests/api_weight.cc: Extend checkstatsweight1 to check that Weight::get_collection_freq() returns the same number as Database::get_collection_freq(). Tue Jul 02 05:00:30 GMT 2013 Olly Betts <olly@survex.com> * api/omenquire.cc,api/omenquireinternal.h, backends/remote/remote-database.cc,backends/remote/remote-database.h, common/remoteprotocol.h,configure.ac,docs/remote_protocol.rst, include/xapian/enquire.h,matcher/multimatch.cc,matcher/multimatch.h, net/remoteserver.cc,tests/api_postingsource.cc: Add Enquire::set_time_limit() method which sets a timelimit after which check_at_least will be disabled. Mon Jul 01 23:42:33 GMT 2013 Olly Betts <olly@survex.com> * tests/Makefile.am: Fix to actually removed cached databases before "make check". Mon Jul 01 22:21:41 GMT 2013 Olly Betts <olly@survex.com> * NEWS,api/queryinternal.cc,matcher/valuestreamdocument.h: Revert changes accidentally included in r17373. Sat Jun 29 12:39:10 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/unicode.h,unicode/tclUniData.cc: Change Xapian::Unicode::Internal::get_character_info() to simply return Xapian::Unicode::UNASSIGNED for character outside unicode (ch >= 0x110000) which avoids a special check for such values before we call get_character_info(). Sat Jun 29 09:31:12 GMT 2013 Olly Betts <olly@survex.com> * tests/api_serialise.cc: Remove code from registry* testcases which tries to test the consequences of throwing an exception from a destructor - it's complex to ensure we don't leak memory while doing this (it seems GCC doesn't release the object in this case, but clang does), and it's generally frowned upon, plus C++11 makes destructors noexcept by default. Thu Jun 27 02:36:02 GMT 2013 Olly Betts <olly@survex.com> * weight/weightinternal.h: Initialise total_term_count to zero. Fixes failure of qp_flag_wildcard3 under valgrind in buildbot. Wed Jun 26 05:48:21 GMT 2013 Olly Betts <olly@survex.com> * backends/dbfactory.cc: '# if defined' -> '#ifdef' to match surrounding code better. Wed Jun 26 05:47:38 GMT 2013 Olly Betts <olly@survex.com> * backends/dbfactory.cc: Fix 'unused label' warning when chert backend is disabled. Wed Jun 26 05:40:29 GMT 2013 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Fix previous change to compile. Wed Jun 26 05:34:25 GMT 2013 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Change qp_scale1 to time 5 repetitions of the large query to help average out variations. Wed Jun 26 04:52:33 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Tweak to simplify diff against similar code in xapian-omega. Wed Jun 26 03:56:14 GMT 2013 Olly Betts <olly@survex.com> * api/matchspy.cc: Fix get_description() to work on a ValueCountMatchSpy with NULL internals - fixes matchspy5 testcase with remote backends when configured with --enable-log. * api/matchspy.cc: Add assertions that internal is non-NULL before using it. * include/xapian/matchspy.h: If internal is NULL, make ValueCountMatchSpy::get_total() return 0 rather than dereferencing NULL. Wed Jun 26 03:10:58 GMT 2013 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc,tests/harness/testsuite.h: If -v/--verbose is specified more than once, show the diagnostic output for passing tests as well as failing/skipped ones. Wed Jun 26 03:08:08 GMT 2013 Olly Betts <olly@survex.com> * tests/api_scalability.cc: Change querypairwise1_helper to repeat the query build 100 times, as with a fast modern machine we were sometimes trying to so many subqueries that we would run out of stack. Wed Jun 26 02:55:20 GMT 2013 Olly Betts <olly@survex.com> * weight/tfidfweight.cc: Wrap long lines. Wed Jun 26 02:00:12 GMT 2013 Olly Betts <olly@survex.com> * configure.ac,m4-macros/xapian-1.3.m4: Fix stripping of _git suffix from snapshot versions not to relying of sed having support for \| as it doesn't on OS X. Tue Jun 25 13:21:44 GMT 2013 Olly Betts <olly@survex.com> * weight/tfidfweight.cc: Merge "impossible" cases in switch statements into possible ones which contain the same code, adding assertions that the "impossible" hasn't happened. More robust and improves test coverage. Tue Jun 25 06:28:00 GMT 2013 Olly Betts <olly@survex.com> * m4-macros/xapian-1.3.m4: XO_LIB_XAPIAN now handles the user specifying a relative path in XAPIAN_CONFIG, e.g.: "./configure XAPIAN_CONFIG=../xapian-core/xapian-config" Tue Jun 25 05:34:41 GMT 2013 Olly Betts <olly@survex.com> * m4-macros/xapian-1.3.m4: Fix comment typo. Tue Jun 25 05:33:54 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Use AC_PATH_PROG instead of AC_PATH_PROG when there's only one name for the program to look for. Sun Jun 23 13:12:54 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: clang doesn't support -Wstrict-null-sentinel or -Wlogical-op, so don't pass it these options. Sat Jun 22 13:40:40 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc,backends/chert/chert_compact.cc: Fix failure when built with -D_GLIBCXX_DEBUG - we were modifying the top() element of a heap before calling pop(), such that the heap comparison operation (which is called when -D_GLIBCXX_DEBUG is on to verify the heap is valid) would read off the end of the data. In a normal build, this issue would likely never manifest. Sat Jun 22 11:21:22 GMT 2013 Olly Betts <olly@survex.com> * common/bitstream.cc: Fix assertion in BitReader::decode(). Sat Jun 22 09:13:13 GMT 2013 Olly Betts <olly@survex.com> * matcher/localsubmatch.cc: Fix assertion failure when built with --enable-assertions. The behaviour when built without assertions happened to be correct. Fri Jun 21 13:10:53 GMT 2013 Olly Betts <olly@survex.com> * tests/unittest.cc: Add simple test for log2(). Fri Jun 21 13:09:09 GMT 2013 Olly Betts <olly@survex.com> * common/Makefile.mk,common/log2.h,configure.ac: Add log2.h header which provides an implementation of log2() if it isn't available in the standard libraries. Fri Jun 14 13:20:58 GMT 2013 Olly Betts <olly@survex.com> * common/fd.h: Add missing header include guards. Fri Jun 14 13:19:44 GMT 2013 Olly Betts <olly@survex.com> * common/append_filename_arg.h,geospatial/geoencode.h: Fix header include guard names which don't match our naming conventions. Fri Jun 14 13:18:40 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_dbcheck.h,backends/chert/chert_dbcheck.h, backends/multi/multi_alltermslist.h: Fix header include guard macro names which don't match the source file names. Fri Jun 14 12:48:14 GMT 2013 Olly Betts <olly@survex.com> * common/compression_stream.h: Whitespace tweak. Thu Jun 13 13:58:14 GMT 2013 Olly Betts <olly@survex.com> * backends/chert/chert_positionlist.cc: Remove 'Assert(rd);' since rd is no longer a pointer. Thu Jun 13 13:28:43 GMT 2013 Olly Betts <olly@survex.com> * tests/harness/utestsuite.cc: Add missing GPL boilerplate and copyright statements. Thu Jun 13 13:26:07 GMT 2013 Olly Betts <olly@survex.com> * common/compression_stream.cc,common/compression_stream.h: Add missing GPL boilerplate and copyright statements. Thu Jun 13 13:24:34 GMT 2013 Olly Betts <olly@survex.com> * common/compression_stream.cc,common/compression_stream.h: Remove unused header includes; don't use "using namespace std;" in a header. Thu Jun 13 08:30:26 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc: Remove 'Assert(rd);' since rd is no longer a pointer. Thu Jun 13 05:07:27 GMT 2013 Olly Betts <olly@survex.com> * common/compression_stream.h: Update include guard to match source file name. Thu Jun 13 02:38:20 GMT 2013 Olly Betts <olly@survex.com> * common/,net/length.h,net/serialise.h: Fix @file markup which wasn't actually in a doxygen comment. Thu Jun 13 02:14:22 GMT 2013 Olly Betts <olly@survex.com> * preautoreconf: Remove needless use of perl prototype. Tue Jun 11 00:40:37 GMT 2013 Olly Betts <olly@survex.com> * common/remoteprotocol.h,docs/remote_protocol.rst: Bump remote protocol version to 38.o. Tue Jun 11 00:37:01 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * api/leafpostlist.cc,matcher/,net/serialise.cc,weight/weight.cc, weight/weightinternal.cc,weight/weightinternal.h: Make collection frequency available to Xapian::Weight subclasses. Extracted from Gaurav Arora's LM branch. Wed May 22 10:04:25 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: When generating ABI compatibility checks in xapian/version.h, pass $CXXFLAGS and $CPPFLAGS to $CXXCPP as they could contain options which affect the ABI (such as -fabi-version for GCC). (Fixes #622) Wed May 22 04:20:39 GMT 2013 Olly Betts <olly@survex.com> * common/win32_uuid.cc: Microsoft GUIDs in binary form have reversed byte order in the first three components compared to standard UUIDs, so the same database would report a different UUID on Windows to on other platforms. With this fix, the UUIDs of existing databases will appear to change on Windows (except in rare "palindronic" cases). Fri May 17 05:52:43 GMT 2013 Olly Betts <olly@survex.com> * docs/overview.rst: Correct documentation for stub files specifying "prog" remote databases. Thu May 16 05:47:40 GMT 2013 Olly Betts <olly@survex.com> * backends/chert/chert_dbcheck.cc: Code tweak for clarity. Thu May 16 05:46:42 GMT 2013 Olly Betts <olly@survex.com> * backends/chert/chert_dbcheck.cc: Fix database check that first docid in each doclength chunk is more than the last docid in the previous chunk - previously this didn't actually work. Thu May 16 05:45:04 GMT 2013 Olly Betts <olly@survex.com> * common/bitstream.h: Fix database check not to falsely report "position table: Junk after position data" whenever there are 7 unused bits (7 is OK, *more* than 7 isn't). Tue May 14 12:42:11 GMT 2013 Olly Betts <olly@survex.com> * m4-macros/xapian-1.3.m4: Adjust XO_LIB_XAPIAN to strip _gitNNN suffix from snapshot versions. Sun May 12 06:11:36 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Simplify previous change. Sun May 12 06:06:47 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Handle git snapshot naming when calculating REVISION. Wed May 08 11:56:27 GMT 2013 Olly Betts <olly@survex.com> * HACKING,INSTALL,Makefile.am,README,docs/Makefile.am,docs/install.rst, docs/stemming.rst,languages/Makefile.mk,queryparser/Makefile.mk, tests/Makefile.am,tests/perftest/perftest.cc: SVN -> git. Wed May 08 08:51:22 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_version.cc,backends/chert/chert_version.cc: Remove unused '#include <cstdio>'. Mon May 06 11:10:25 GMT 2013 Olly Betts <olly@survex.com> * api/omenquire.cc,configure.ac,include/xapian/enquire.h: Merge two Enquire::get_eset() overloaded forms into one with a default value for parameter min_wt. Thu May 02 12:05:28 GMT 2013 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update for 1.3.1. Thu May 02 11:13:30 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/weight.h,tests/api_nodb.cc,tests/api_weight.cc, weight/tfidfweight.cc: Use lowercase letters for the TfIdfWeight descriptions to match what SMART uses. Wed May 01 05:13:07 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/database.h: Work around Apple's OS X SDK defining a check() macro. Tue Apr 30 13:51:53 GMT 2013 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Mon Apr 29 03:09:16 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_dbcheck.cc,backends/chert/chert_dbcheck.cc: Use decode_interpolative_next() when checking database consistency. We also now complain about any junk after the position data. * common/bitstream.cc,common/bitstream.h: Remove the old version of decode_interpolative() which unpacks into a std::vector. Mon Apr 29 02:16:54 GMT 2013 Olly Betts <olly@survex.com> * backends/chert/chert_positionlist.cc, backends/chert/chert_positionlist.h: Backport positionlist changes from brass. Mon Apr 29 01:58:59 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc: No need to initialise the BitReader object in the single-entry position list case. Mon Apr 29 01:58:10 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc: Fix bug when reaching the last positionlist entry in skip_to(). Mon Apr 29 00:46:10 GMT 2013 Olly Betts <olly@survex.com> * tests/api_posdb.cc: Fix comment typos. Mon Apr 29 00:27:42 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc: Fix handling of single-entry position lists. Sun Apr 28 23:45:02 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc: Fix BrassPositionList for case when there's no positional information for a term. Sun Apr 28 10:29:59 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc: Simplify setting of have_started in BrassPositionList::skip_to(). Sun Apr 28 10:19:52 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc, backends/brass/brass_positionlist.h,common/bitstream.h: Make the BitReader object a member of BrassPositionList rather than tracking a pointer to it. Sun Apr 28 07:22:07 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc,common/bitstream.h: Reuse the same BitReader object for each document rather than deleting the old one and creating a new one. Sun Apr 28 07:06:12 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc: Move check for reaching the end after the next() in the loop in skip_to(), as we know we aren't at the end on the first iteration, and we've now handled the case of termpos == last before we get here. Sun Apr 28 07:04:54 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc: We know what the last entry in the position list is, so special case skip_to(last_entry) to just move there rather than calling next() a lot of times. Sun Apr 28 07:03:57 GMT 2013 Olly Betts <olly@survex.com> * common/bitstream.h: We don't need to stack pos_j when doing lazy interpolative decoding. Sun Apr 28 06:48:01 GMT 2013 Olly Betts <olly@survex.com> * common/bitstream.h: Store j and k before pos_j and pos_k as it's fractionally more efficient. Sun Apr 28 06:44:32 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc, backends/brass/brass_positionlist.h,common/bitstream.cc, common/bitstream.h: Hook up BrassPositionList to the new lazy interpolative decoder. Fri Apr 26 02:09:38 GMT 2013 Marius Tibeica <mtibeica@gmail.com> * common/bitstream.cc,common/bitstream.h: Support reading entries from an interpolative coded list one by one. Thu Apr 25 10:57:02 GMT 2013 Olly Betts <olly@survex.com> * docs/remote_protocol.rst: Correct error in documentation of REPLY_DOCDATA message. Thu Apr 25 10:44:56 GMT 2013 Marius Tibeica <mtibeica@gmail.com> * common/bitstream.cc: renamed my_fls to highest_order_bit. Thu Apr 25 06:49:48 GMT 2013 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remoteprotocol.h, docs/remote_protocol.rst,net/remoteserver.cc: Prefix compress list of terms and metadata keys in the remote protocol. This requires a remote protocol major version bump. Wed Apr 24 23:14:47 GMT 2013 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Fix comment typos. Wed Apr 24 06:27:26 GMT 2013 Olly Betts <olly@survex.com> * include/xapian.h: Add catch for Wt defining a macro called slots, like we already do for Qt. Wed Apr 24 03:40:13 GMT 2013 Olly Betts <olly@survex.com> * NEWS: Update. Tue Apr 23 12:34:23 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc: Use a block's revision number to check if it was newly allocated in the current revision, and assert that the base file shows it block_free_at_start (previously the check and assertion were the other way around). Tue Apr 23 12:11:21 GMT 2013 Olly Betts <olly@survex.com> * HACKING: Note macports needed for development work. Tue Apr 23 11:41:17 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: Fix typo in doc comment. Tue Apr 23 11:39:30 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: Disable error for direct inclusion of xapian/version.h for now. Sun Apr 21 07:49:38 GMT 2013 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: Fix quotes in new #error in version.h. Sun Apr 21 06:25:57 GMT 2013 Olly Betts <olly@survex.com> * configure.ac,generate-exceptions,include/xapian/: Give an error if user code tries to include API headers other than xapian.h directly. Sat Apr 20 10:12:23 GMT 2013 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Generate '#include <config.h>' at the start of the stemmer code. Fri Apr 19 23:36:25 GMT 2013 Olly Betts <olly@survex.com> * configure.ac: Remove empty default cases. Wed Apr 17 03:06:57 GMT 2013 Olly Betts <olly@survex.com> * NEWS: Update from 1.2.15 and ChangeLog. Thu Apr 11 22:17:44 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_version.cc,backends/chert/chert_version.cc: Call io_sync() on the version file when we create it. Tue Apr 09 11:02:35 GMT 2013 Olly Betts <olly@survex.com> * bin/xapian-delve.cc: If -v is specified more than once, show even more info in some cases. Tue Apr 09 09:40:02 GMT 2013 Olly Betts <olly@survex.com> * tests/Makefile.am: Force automake to use the serial test driver, which requires automake 1.12 (to recognise the serial-tests driver). Mon Apr 08 06:30:40 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * weight/: Added tfidfweight.cc containing the implementation of the TfIdfWeight class. * include/xapian/weight.h: Added TfIdfWeight class for the tf-idf weighting scheme. * tests/api_weight.cc: Added tests for TfIdfWeight. * tests/api_nodb.cc: Added simple tests for TfIdfWeight. Sat Apr 06 00:21:34 GMT 2013 Olly Betts <olly@survex.com> * m4-macros/xapian-1.3.m4: XO_LIB_XAPIAN now handles the user specifying XAPIAN_CONFIG without a path - e.g.: "./configure XAPIAN_CONFIG=xapian-config-1.3" Thu Apr 04 12:02:36 GMT 2013 Olly Betts <olly@survex.com> * common/Makefile.mk: Remove random comment. Thu Apr 04 10:06:53 GMT 2013 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,tests/Makefile.am: Remove support for 'configure --enable-quiet', 'make QUIET=' and 'make QUIET=y' - automake now supports 'configure --enable-silent-rules', 'make V=1' and 'make V=0' which are broadly equivalent and more standard. Thu Apr 04 09:58:35 GMT 2013 Olly Betts <olly@survex.com> * HACKING: Update to note that automake 1.12.2 is now a hard requirement. Tue Apr 02 10:22:44 GMT 2013 Olly Betts <olly@survex.com> * queryparser/lemon.c: Comment typo fixes. Tue Apr 02 09:47:14 GMT 2013 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Sync with latest upstream version of lemon. The only notable change is a bug fix for platforms where sizeof(long) < sizeof(void*) (LLP64 model). Tue Apr 02 05:51:12 GMT 2013 Olly Betts <olly@survex.com> * HACKING: Tell vim to syntax highlight as rst. Tue Apr 02 05:50:26 GMT 2013 Olly Betts <olly@survex.com> * HACKING: Update for recent changes to bootstrap. In particular, we now bootstrap snapshots with automake 1.13.1. Fri Mar 29 05:18:11 GMT 2013 Olly Betts <olly@survex.com> * net/tcpclient.cc: Fix __WIN32__ conditional, accidentally broken in last change to this file. Thu Mar 28 04:19:26 GMT 2013 jiangwen jiang <jiangwen127@gmail.com> * api/expanddecider.cc,include/xapian/expanddecider.h, tests/api_anydb.cc,tests/api_nodb.cc: Add ExpandDeciderFilterPrefix class to only return terms with a particular prefix. (fixes #467) Thu Mar 21 23:55:26 GMT 2013 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add test coverage for change to QueryParser CJK handling in previous commit. Thu Mar 21 07:21:30 GMT 2013 Greg Banks <gnb@fastmail.fm> * queryparser/cjk-tokenizer.cc,queryparser/termgenerator_internal.cc, tests/termgentest.cc: Some CJK codepoints are also punctuation and hence not word characters, and should not really be used as fodder for generating N-grams. Also fixes a pre-existing bug where every second sequence of N-grammable CJK characters in the text failed to generate the first 2-gram. Sun Mar 17 20:45:48 GMT 2013 Olly Betts <olly@survex.com> * common/pretty.h: Restore space after comma in output. Sun Mar 17 03:59:52 GMT 2013 Dan Colish <dcolish@gmail.com> * common/pretty.h: Correct recursion issue. Sun Mar 17 02:58:52 GMT 2013 Dan Colish <dcolish@gmail.com> * common/pretty.h: Resolve ambiguity by moving operator| to a member function of PrettyOStream Sat Mar 16 21:43:32 GMT 2013 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: std::less provides a total order over pointers, so we can just use that to implement filter_group_id::operator<() - we only need a consistent order so we can use it as the key to a std::map. Fri Mar 15 06:36:32 GMT 2013 Dan Colish <dcolish@gmail.com> * backends/brass/brass_table.cc,backends/chert/chert_table.cc, common/pretty.h: Avoid needlessly casting away const-ness in debug logging. Fri Mar 15 06:31:32 GMT 2013 Olly Betts <olly@survex.com> * matcher/multiandpostlist.cc,matcher/multixorpostlist.cc: Add a couple more LOGCALL annotations. Fri Mar 15 06:27:42 GMT 2013 Dan Colish <dcolish@gmail.com> * api/registry.cc,common/pretty.h: Fix incorrect LOGCALL return value. Fri Mar 15 06:24:11 GMT 2013 Olly Betts <olly@survex.com> * api/registry.cc,common/Makefile.mk,common/registryinternal.h: common/registryinternal.h is only included by api/registry.cc so just inline its contents. Fri Mar 15 06:11:14 GMT 2013 Olly Betts <olly@survex.com> * api/registry.cc: No need to explicitly default initialise std::map members or the base class. Fri Mar 08 04:05:31 GMT 2013 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc,backends/chert/chert_compact.cc: Fix multipass compaction not to damage document values, and to merge the database stats correctly. (fixes #615) * tests/api_compact.cc: Add regression test compactmultipass1. Thu Mar 07 00:34:46 GMT 2013 Aarsh Shah <aarshkshah1992@gmail.com> * tests/api_anydb.cc: Added tradweight4 testcase to test TradWeight with an RSet for weighting documents. Wed Mar 06 08:55:12 GMT 2013 Olly Betts <olly@survex.com> * HACKING: gs package has been replaced by ghostscript package in current Debian and Ubuntu releases. Tue Feb 19 04:17:19 GMT 2013 Olly Betts <olly@survex.com> * common/Tokeniseise.pm: Add the ability to append lines to the header. Fri Feb 15 05:37:09 GMT 2013 Olly Betts <olly@survex.com> * common/realtime.h: Fix RealTime::sleep() to calculate the delta to the time we want to sleep to the right way round. Previously we wouldn't sleep if we were before the deadline, but would sleep if the deadline had already passed (and by the amount the deadline had passed)! And on Windows, fix the sleep to be for the delta rather than 43+ years (the time elapsed since the start of 1970). (Fixes #472) Fri Feb 15 04:09:28 GMT 2013 Olly Betts <olly@survex.com> * api/queryinternal.cc: Need <functional> for mem_fun(). Fri Feb 15 03:56:32 GMT 2013 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Don't provide explicit template types to make_pair - it isn't useful, and breaks with C++11. Fixes build error with MSVC2012. Thu Jan 31 23:39:08 GMT 2013 Olly Betts <olly@survex.com> * common/Tokeniseise.pm: Put the binary chop tables with the larger ones last, so the offsets to the start of them are more likely to fit into a byte. Wed Jan 30 04:17:07 GMT 2013 Olly Betts <olly@survex.com> * common/Makefile.mk,common/Tokeniseise.pm,languages/Makefile.mk, languages/collate-sbl: Factor out the generic code to create a string-to-enum mapping table into a perl module so we can reuse it elsewhere. Mon Jan 28 01:44:53 GMT 2013 Olly Betts <olly@survex.com> * examples/quest.cc: Fix to build with Sun Studio 12 compiler. (ticket#611) Thu Jan 24 22:40:12 GMT 2013 Olly Betts <olly@survex.com> * common/Makefile.mk,common/keyword.cc,common/keyword.h: Commit new keyword lookup function needed by previous commit. Thu Jan 24 22:26:19 GMT 2013 Olly Betts <olly@survex.com> * languages/: Generate a compact and efficient table to convert language names to enum codes, which we can then use a C switch statement to dispatch. The table first checks the token length, and then does a binary chop on tokens of the same length. This is both faster and smaller than the approach we were using, with the benefit that the table is auto-generated. Wed Jan 23 00:51:42 GMT 2013 Olly Betts <olly@survex.com> * matcher/selectpostlist.cc,matcher/selectpostlist.h: Check document matches weight threshold before checking positional information, which speeds up slow phrase searches (tweaked version of positional-query-weight-check-first.patch from #394). Thu Jan 17 06:48:41 GMT 2013 Olly Betts <olly@survex.com> * tests/api_stem.cc: Extend stem2 to ensure description of the stemmer isn't the same as the description of Stem("none"). Fix description of testcase stemlangs2. Thu Jan 17 02:32:31 GMT 2013 Olly Betts <olly@survex.com> * HACKING: Update link to new home of autotools tutorial. Wed Jan 16 02:40:20 GMT 2013 Olly Betts <olly@survex.com> * HACKING,Makefile.am: Rename CHK_SOURCES to check_sources - variables ending _SOURCES usually have a special meaning in automake, so it's confusing to have one which doesn't. Wed Jan 16 01:55:50 GMT 2013 Olly Betts <olly@survex.com> * api/omenquire.cc: Don't cache documents retrieved by MSet::get_document() unless they were requested with fetch(). This avoids using a lot of memory when many MSet entries are retrieved. (Fixes #604) Wed Jan 09 11:49:58 GMT 2013 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog and 1.2 branch. Fri Jan 04 23:14:33 GMT 2013 Olly Betts <olly@survex.com> * net/progclient.cc: Fix typo so we actually use SOCK_CLOEXEC. Fri Jan 04 23:14:01 GMT 2013 Olly Betts <olly@survex.com> * common/safesyssocket.h: We need the same workaround for socket() too. Fri Jan 04 04:41:17 GMT 2013 Olly Betts <olly@survex.com> * common/safesyssocket.h: On Linux at least, sometimes SOCK_CLOEXEC is defined but the kernel doesn't handle it in socketpair(), so add a wrapper which will retry without SOCK_CLOEXEC in this case. Noticed on Ubuntu precise. Sat Dec 29 17:53:30 GMT 2012 Olly Betts <olly@survex.com> * docs/quickstart.rst: Fix seriously outdated statement that Xapian doesn't create the database directory - that changed in 0.7.2 (release 2003-07-11), pointed out by aarsh on #xapian. Thu Dec 27 06:06:30 GMT 2012 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Adjust handling of Unicode opening/closing double quotes - if a quoted boolean term was started with ASCII double quote, then only ASCII double quote can end it, as otherwise it's impossible to quote a term containing Unicode double quotes. * tests/queryparsertest.cc: Update testcases. Sun Dec 23 18:16:52 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: Fix typo in previous commit. Sun Dec 23 18:06:45 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: Use the new flock()-based flint-compatible locking under DJGPP, and remove special cases to disable chert and brass under DJGPP (untested, as I no longer have a DJGPP setup or cross-compiler). Sun Dec 23 17:51:36 GMT 2012 Olly Betts <olly@survex.com> * backends/flint_lock.cc,backends/flint_lock.h: Add flock() based locking implementation, which is much simpler than using fcntl() due to saner semantics around releasing locks when closing other descriptors on the same file (at least on platforms where flock() isn't just a compatibility wrapper around fcntl()). We can't simply switch to this without breaking locking compatibility with previous releases, though it might be useful for porting to platforms without fcntl() locking. Also, flock() apparently doesn't work over NFS - perhaps that's OK, but we should at least check the failure mode. Sat Dec 22 06:29:23 GMT 2012 Olly Betts <olly@survex.com> * backends/flint_lock.cc: Only set close-on-exec on the lockfile fd if we are able to clear it in the child process, and clear it in the child process if we set it when we opened it. Sat Dec 22 06:14:40 GMT 2012 Olly Betts <olly@survex.com> * backends/flint_lock.cc,net/progclient.cc: Use SOCK_CLOEXEC with socketpair(), and then clear the close-on-exec flag for child process end of the pair before we call exec. Sat Dec 22 06:13:44 GMT 2012 Olly Betts <olly@survex.com> * net/progclient.cc: Remove O_CLOEXEC where we open stderr on /dev/null before calling exec as we want that fd to survive the exec! Sat Dec 22 05:53:12 GMT 2012 Olly Betts <olly@survex.com> * HACKING,backends/flint_lock.cc,common/Makefile.mk, common/safesyssocket.h,net/progclient.cc,net/tcpclient.cc, net/tcpserver.cc,tests/harness/backendmanager_remotetcp.cc: Use SOCK_CLOEXEC where available, and try FD_CLOEXEC if SOCK_CLOEXEC isn't available. Sat Dec 22 04:09:23 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/,backends/chert/,backends/flint_lock.cc, common/debuglog.cc,common/replicate_utils.cc,common/safefcntl.h, net/progclient.cc,net/remoteconnection.cc, tests/harness/testsuite.cc: If O_CLOEXEC is supported, pass it to open() so fds we open are closed if the application using us calls exec(). Fri Dec 21 03:48:36 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: Don't autodisable the remote backend if we fail to find a required function. It's more likely the user needs to pass LIBS=-lfoo and it's more helpful to error out so they can do that, or pass --disable-backend-remote if they really don't want it, rather than having them not notice the automatic disabling until after they've built. This also matches what we do for the disk based backends. Fri Dec 21 03:48:11 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: Expand comment about DJGPP. Fri Dec 21 03:47:13 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: Update comment - we've supported the prog backend on Windows for more than 5 years now! Fri Dec 21 03:32:13 GMT 2012 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Use O_BINARY in replication tests. Fri Dec 21 03:24:37 GMT 2012 Olly Betts <olly@survex.com> * common/replicate_utils.cc: Make posixy_open() always use O_BINARY. * backends/brass/brass_btreebase.cc,backends/brass/brass_database.cc, backends/brass/brass_databasereplicator.cc, backends/chert/chert_btreebase.cc, backends/chert/chert_databasereplicator.cc,common/posixy_wrapper.cc: Don't pass O_BINARY to posixy_open() in the places where we were (which wasn't everywhere, probably causing replication bugs under __WIN32__). Fri Dec 21 02:57:56 GMT 2012 Olly Betts <olly@survex.com> * net/remoteconnection.cc,net/remoteconnection.h: Only "outline" RemoteConnection dtor under __WIN32__. Fri Dec 21 02:56:28 GMT 2012 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc: Make the UnimplementedError for a MatchSpy which doesn't implement name() clearer that it's this particular subclass which can't be used remotely, not all MatchSpy objects. Thu Dec 20 22:27:40 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: Use user-specified LIBS for configure tests, which is what's expected, and provides a way for the user to tell configure where to find library functions which configure can't find for itself. Thu Dec 20 22:07:19 GMT 2012 Olly Betts <olly@survex.com> * api/compactor.cc,api/replication.cc,backends/brass/,backends/chert/, common/,net/remoteconnection.cc: Rename msvc_posix_rename(), etc to posixy_rename() (avoid a posix_ prefix as that's used by some library functions, and these aren't necessarily 100% POSIX compatible), and provide macro versions on Unix, so we can avoid a #ifdef at every call site. Thu Dec 20 20:57:18 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: socketpair() needs -lnetwork on Haiku. Thu Dec 20 02:30:12 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_databasereplicator.cc, backends/chert/chert_databasereplicator.cc: Simplify how we open the .DB file on the replication slave to just call open() once with O_CREAT, rather than once without, than stat() if that fails, and then again with O_CREAT|O_TRUNC if stat() doesn't show an ordinary file exists. Thu Dec 20 02:05:12 GMT 2012 Olly Betts <olly@survex.com> * api/compactor.cc,common/,languages/compiler/syswords2.h, tests/api_compact.cc: Remove extra blank lines at end of files. Thu Dec 20 01:54:47 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc,backends/chert/chert_database.cc: Fix "if (fd > 0)" tests to be "if (fd >= 0)" in code related to replication. In practice this is unlikely to actually have caused problems in real world cases. Sun Dec 16 23:10:13 GMT 2012 Olly Betts <olly@survex.com> * bin/Makefile.mk: INCLUDES is always used, AM_CPPFLAGS isn't if there are per-executable CPPFLAGS, so add AM_CPPFLAGS explicitly when we use per-executable CPPFLAGS. Fri Dec 14 21:25:58 GMT 2012 Olly Betts <olly@survex.com> * Makefile.am,languages/Makefile.mk,queryparser/Makefile.mk, tests/Makefile.am: In automake, INCLUDES is now deprecated in favour of AM_CPPFLAGS so update to use the latter. Tue Nov 27 05:39:30 GMT 2012 Olly Betts <olly@survex.com> * examples/quest.cc: Need <algorithm> for lower_bound(). Mon Nov 26 23:46:43 GMT 2012 Olly Betts <olly@survex.com> * examples/quest.cc: Align option descriptions in --help output, and make the initial letter of such descriptions consistently lowercase. Mon Nov 26 19:52:50 GMT 2012 Olly Betts <olly@survex.com> * examples/quest.cc: Add --flags command line option to allow setting arbitrary QueryParser flags. Mon Nov 26 03:02:03 GMT 2012 Olly Betts <olly@survex.com> * api/valuerangeproc.cc: Move 3 declarations to where we first use the variables. Mon Nov 26 02:59:41 GMT 2012 Olly Betts <olly@survex.com> * matcher/multimatch.cc: We no longer use the highest weighted MSet entry to calculate percentages, so stop searching for it. (Spotted by cppcheck) Mon Nov 26 02:59:02 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc,backends/chert/chert_compact.cc: Remove unused variables (spotted by cppcheck). Mon Nov 26 02:12:19 GMT 2012 Olly Betts <olly@survex.com> * api/maptermlist.h,api/matchspy.cc,api/omdatabase.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_positionlist.cc, backends/multi/multi_postlist.cc,backends/remote/net_termlist.cc, bin/xapian-delve.cc,matcher/exactphrasepostlist.cc, matcher/mergepostlist.cc,matcher/phrasepostlist.cc, tests/harness/backendmanager.cc: Prefer ++i to i++. Fri Nov 23 03:38:39 GMT 2012 Olly Betts <olly@survex.com> * HACKING,Makefile.am: With lcov 1.10, we no longer need any of our hacky workarounds, so make this a requirement and simplify the rules in the makefile. Thu Nov 22 23:34:09 GMT 2012 Olly Betts <olly@survex.com> * Makefile.am: Pass -f to lcov, which should avoid the need for the patch we've been applying to lcov. Thu Nov 22 01:34:05 GMT 2012 Olly Betts <olly@survex.com> * backends/flint_lock.cc: Close excess file handles before we get the fcntl() lock, just in case one of those file handles is open on the lock file. * tests/api_backend.cc: Add regression test lockfilealreadyopen1. Thu Nov 22 00:35:51 GMT 2012 Olly Betts <olly@survex.com> * common/closefrom.cc,configure.ac: On platforms with the F_MAXFD fcntl but without closefrom(), we were failing to close the highest numbered open fd in our closefrom() replacement. We now also work around valgrind not hiding some extra fds it has open, but then complaining if we try to close them. Wed Nov 21 22:50:21 GMT 2012 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: When reporting valgrind error, skip any warnings. Tue Nov 20 07:26:24 GMT 2012 Olly Betts <olly@survex.com> * tests/api_anydb.cc: msetweights1 makes multidb1 and multidb2 redundant, so remove them. Tue Nov 20 07:21:46 GMT 2012 Olly Betts <olly@survex.com> * tests/api_backend.cc: Extend msetweights1 to test the case where only even docids match, so in the multi_* case, only docids from one subdatabase match. Tue Nov 20 07:04:55 GMT 2012 Olly Betts <olly@survex.com> * tests/api_backend.cc: Add testcase msetweights1 which checks we get the weights we expect for a two term query. Mon Nov 19 21:17:33 GMT 2012 Olly Betts <olly@survex.com> * queryparser/termgenerator.cc: TermGenerator internal member can never be NULL, so drop NULL test. Sat Nov 17 20:54:50 GMT 2012 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Remove compatibility code for valgrind < 3.3.0, since we reject such old versions at configure time. Sat Nov 17 00:05:36 GMT 2012 Olly Betts <olly@survex.com> * matcher/localsubmatch.cc: Optimise an unweighted query term which matches all the documents in a subdatabase to use the "MatchAll" postlist. (ticket#387) Fri Nov 16 23:36:34 GMT 2012 Olly Betts <olly@survex.com> * tests/api_weight.cc: New testcase checkstatsweight1 which checks a weight subclass gets the correct values for all the stats. Fri Nov 16 07:30:18 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: Generate a check for compatible _DEBUG settings if built with MSVC. (ticket#389) Fri Nov 16 01:28:55 GMT 2012 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Improve a couple of comments. Fri Nov 16 01:25:28 GMT 2012 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Implicitly close any unclosed brackets at the end of the query string. Patch from Sehaj Singh Kalra. * tests/queryparsertest.cc: Adjust existing testcases and feature tests. Thu Nov 15 23:57:10 GMT 2012 Olly Betts <olly@survex.com> * tests/api_query.cc: Add testcase xor3 to ensure that XOR handles all remaining subqueries running out at the same time. Thu Nov 15 10:57:32 GMT 2012 Olly Betts <olly@survex.com> * tests/api_query.cc: One test here does need a backend, and it seems a more useful category, so drop the "don't need a backend" part from the @brief documentation comment. Thu Nov 15 09:29:47 GMT 2012 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix calculation of 0.0/0.0 in some cases. This then got used as a minimum weight, but it seems this gives -nan (at least on x86-64 Linux) so it may have been harmless in practice. Thu Nov 15 05:39:47 GMT 2012 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Eliminate workarounds for egcs 1.1.2 (certainly we no longer support a compiler that old, but maybe some newer compilers have the same issue, so I've rewritten the code to iterate backwards over one MSet, which is a useful feature to have coverage for). Thu Nov 15 05:28:40 GMT 2012 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Fix reversebool1 and reversebool2 to actually use the result when they check an MSetIterator against end(). Thu Nov 15 01:55:15 GMT 2012 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc: Add --quiet option to xapian-replicate, and be a little more verbose by default. Thu Nov 15 01:54:14 GMT 2012 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc: xapian-replicate --force-copy now only forces a single copy, and then attempts to replicate incrementally. Thu Nov 15 00:51:50 GMT 2012 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc: Tweak code layout. Wed Nov 14 21:40:45 GMT 2012 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Allow files > 32G to be be copied by replication. Tue Nov 13 03:44:03 GMT 2012 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Pass 'true' or 'false' for 'bool' parameter, not '0' or '1' (especially confusing as the two previous parameters were integers which were often 0 or 1...) Tue Nov 13 03:34:42 GMT 2012 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc,net/replicatetcpclient.cc, net/replicatetcpclient.h: Add --full-copy option to xapian-replicate. * tests/api_replicate.cc: New testcase replicate6 which tests the same code path used by 'xapian-replicate --full-copy'. Tue Nov 13 02:55:44 GMT 2012 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Use '++' instead of '+= 1'. Drop superfluous 'return;' at end of function returning void. Tue Nov 13 01:12:33 GMT 2012 Olly Betts <olly@survex.com> * tests/harness/unixcmds.cc: Tweak to avoid default initialisation of string. Mon Nov 12 23:25:14 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/unicode.h: GCC 4.6.3 doesn't optimise the asr idion either. Add links to the PR I filed and the GCC documentation that signed right shift does sign extension. Mon Nov 12 10:21:19 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/unicode.h,unicode/tclUniData.cc: Update Unicode character database to 6.2.0. * tests/api_unicode.cc: Add tests for U+20BA, added in 6.2.0. Sat Nov 10 05:39:16 GMT 2012 Olly Betts <olly@survex.com> * api/leafpostlist.cc,matcher/multiandpostlist.cc: Improved fix for #590 - count all matching LeafPostList objects with a Weight object rather than trying to prune at the MultiAndPostList level based on max_wt (if wdf is always zero for a term, BM25 gives max_wt of 0, which leads to us never counting that subquery. * tests/api_percentages.cc: Add regression test topercent7. Fri Nov 09 04:25:32 GMT 2012 Olly Betts <olly@survex.com> * tests/api_percentages.cc: Note bug number in regression test for #590. Tue Nov 06 22:42:31 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc: Restore two missing lines in database checking where we report a block with the wrong level. Tue Oct 16 11:19:42 GMT 2012 Olly Betts <olly@survex.com> * api/compactor.cc,backends/dbcheck.cc,backends/dbfactory.cc, common/Makefile.mk: Fix to build when configured with --disable-backend-brass --disable-backend-chert. (ticket#586) Mon Oct 15 04:07:30 GMT 2012 Olly Betts <olly@survex.com> * HACKING: automake 1.12.4 is now used to generate snapshots and releases. Sat Oct 13 09:34:21 GMT 2012 Olly Betts <olly@survex.com> * weight/weightinternal.h: Revert addition of "#include <config.h>" in r16820. * HACKING: Document that we don't include <config.h> from header files, and the reason why. Sat Oct 13 09:00:23 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Fix documentation comment typos: "covert" -> "convert". Thu Oct 11 12:28:30 GMT 2012 Olly Betts <olly@survex.com> * docs/valueranges.rst: Update documentation to reflect change in Xapian 1.1.2 - DateValueRangeProcessor and StringValueRangeProcessor now support a prefix or suffix. Mon Oct 01 05:05:54 GMT 2012 Olly Betts <olly@survex.com> * HACKING: Try to make it clearer we're looking for a dual-licence on submitted patches. Tue Sep 25 08:09:08 GMT 2012 Olly Betts <olly@survex.com> * common/Makefile.mk,common/append_filename_arg.h, tests/harness/unixcmds.cc: Split out append_filename_argument() into its own file so it can be used elsewhere. Tue Sep 25 06:46:03 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Mark MSetIterator::at_end() and ESetIterator::at_end() as @private @internal so that they don't appear in the API docs. Tue Sep 25 06:45:07 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Remove FIXMEs about converting MSetIterator and ESetIterator to use Internal classes - as HACKING notes, the current implementation is that way for performance reasons. Tue Sep 25 06:36:37 GMT 2012 Olly Betts <olly@survex.com> * tests/harness/unixcmds.cc: Escape filenames on POSIX platforms by wrapping in single quotes and specially handling single quotes in the filename - previously newlines in filenames got eaten by the shell. Tue Sep 25 06:32:36 GMT 2012 Olly Betts <olly@survex.com> * tests/harness/unixcmds.cc: Protect filenames which start with '-' on MS Windows like we already do on other platforms. The built-in commands generally only interpret '/' as the character introducing a command line option (which we already convert to '\'), but many ported programs understand '-' too/instead. Sat Aug 25 16:09:50 GMT 2012 Dan Colish <dcolish@gmail.com> * include/xapian/enquire.h: Add function to test MSetIterator and ESetIterator exhaustion against the size of the underlying interee since comparison against an empty Itor is not suitable for an exhaustion test. Add FIXME that these Iterators should use an Internal class. Thu Aug 09 21:50:52 GMT 2012 Dan Colish <dcolish@gmail.com> * weight/weightinternal.h: Include config.h since macros are used from it. Wed Aug 08 15:40:38 GMT 2012 Dan Colish <dcolish@gmail.com> * api/queryvector.h: Backout previous fix. When compiling with -O0, increase the stack limit to > 8M when running scalability tests. Wed Aug 08 02:10:21 GMT 2012 Dan Colish <dcolish@gmail.com> * api/queryvector.h: Fix SIGBUS when clearing QueryVectors Wed Aug 01 15:25:25 GMT 2012 Dan Colish <dcolish@gmail.com> * HACKING,Makefile.am,tests/Makefile.am: Fix non-portable warnings from automake, add HACKING doc for check-syntax make target Wed Aug 01 04:54:18 GMT 2012 Dan Colish <dcolish@gmail.com> * Makefile.am,api/Makefile,backends/Makefile,backends/brass/Makefile, backends/chert/Makefile,backends/inmemory/Makefile, backends/multi/Makefile,backends/remote/Makefile,bin/Makefile, common/Makefile,examples/Makefile,expand/Makefile, geospatial/Makefile,include/Makefile,include/xapian/Makefile, languages/Makefile,matcher/Makefile,net/Makefile, queryparser/Makefile,tests/harness/Makefile,unicode/Makefile, weight/Makefile: Add check-syntax target to Makefiles to support editor syntax checks Sun Jul 29 15:32:00 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: Bump LIBRARY_VERSION_INFO so I can package a snapshot cleanly. Tue Jul 24 01:35:17 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.cc, queryparser/queryparser.lemony,queryparser/queryparser_internal.h, tests/queryparsertest.cc: Add FieldProcessor class (ticket#128) - currently marked as an experimental API while we sort out how best to sort out exactly how it interacts with other QueryParser features. Tue Jul 24 01:19:03 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Add extra DateValueRangeProcessor constructor overloaded form so that in DateValueRangeProcessor(1, "date:"), the const char * gets interpreted as std::string rather than bool. * tests/queryparsertest.cc: Add regression test. Thu Jul 19 05:32:04 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_dbcheck.h, backends/inmemory/inmemory_alltermslist.cc,bin/xapian-delve.cc: Fix filenames in comments which mismatch actual name of file. Thu Jul 19 05:27:38 GMT 2012 Olly Betts <olly@survex.com> * queryparser/termgenerator_internal.cc: Remove "FIXME: Add API to allow control of how stemming is used?", as we have now added such an API. Thu Jul 19 05:24:49 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/termgenerator.h,queryparser/termgenerator.cc, queryparser/termgenerator_internal.cc, queryparser/termgenerator_internal.h,tests/termgentest.cc: Add new method TermGenerator::set_max_word_length() to allow this limit to be adjusted by the user. Mon Jul 16 03:34:27 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Clarify that the "reverse" parameter of set_sort_by_relevance_then_value() and set_sort_by_relevance_then_key() only affects the ordering of the value/key part of the sort. Fri Jul 13 04:55:26 GMT 2012 Dan Colish <dcolish@gmail.com> * matcher/multimatch.cc: Remove duplicate import Thu Jul 12 05:13:57 GMT 2012 Olly Betts <olly@survex.com> * common/debuglog.cc: Preserve errno over debug logging calls, so they can safely be added to code which expects errno not to change. Mon Jul 02 05:27:43 GMT 2012 Dan Colish <dcolish@gmail.com> * backends/brass/brass_databasereplicator.cc: Use new/delete to avoid variable length array gcc extension and comply with c++98 Mon Jul 02 05:02:14 GMT 2012 Dan Colish <dcolish@gmail.com> * api/omdatabase.cc,tests/api_wrdb.cc: Prefer container.empty() to container.size() == 0. Sun Jul 01 10:46:19 GMT 2012 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog, 1.2.11 and 1.2.12. Fri Jun 29 15:32:11 GMT 2012 Dan Colish <dcolish@> * tests/harness/testsuite.cc: Add safeunistd.h unconditionally for isatty, required by GCC 4.7 Wed Jun 27 10:45:13 GMT 2012 Olly Betts <olly@survex.com> * AUTHORS: Add GSoC students. Tue Jun 26 08:45:03 GMT 2012 Olly Betts <olly@survex.com> * include/xapian.h: Add check for Qt headers being included before us and defining 'slots' as a macro - if they are, give a clear error advising how to work around this. Thu Jun 14 23:48:15 GMT 2012 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Correction - we don't "create a lock file", we "lock a file". Tue Jun 12 13:08:58 GMT 2012 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Make sure XAPIAN_MAX_CHANGESETS gets unset after testcases which set it, so further testcases don't waste time generating changesets. Tue Jun 12 12:11:51 GMT 2012 Olly Betts <olly@survex.com> * docs/replication.rst: The value of XAPIAN_MAX_CHANGESETS does now actually determine how many changesets we keep. Tue Jun 12 01:20:59 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: Force link_all_deplibs_CXX=no for Solaris. Tue Jun 12 01:08:17 GMT 2012 Olly Betts <olly@survex.com> * xapian-config.in: Don't interpret a missing .la file as meaning that we only have static libraries. Sun Jun 10 12:57:16 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_btreebase.cc,backends/brass/brass_btreebase.h, backends/brass/brass_check.cc: Remove is_empty(), tweak calculate_last_block() to calculate the bit map size correctly in the case of an empty table, and use the bit map size instead of is_empty() in xapian-check. * backends/chert/chert_btreebase.cc,backends/chert/chert_btreebase.h, backends/chert/chert_check.cc: Make equivalent changes for chert. Sun Jun 10 11:27:52 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/brass/brass_table.h: We don't need the workaround for existing tables which were built with a 32 bit item count for brass, since that bug was fixed in 1.1.4, which is the same version which brass was added in. Sun Jun 10 11:20:58 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_btreebase.cc,backends/brass/brass_btreebase.h, backends/chert/chert_btreebase.cc,backends/chert/chert_btreebase.h: Remove unused copy constructors for BrassTable_base and ChertTable_base. Wed Jun 06 11:10:26 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Note that STEM_ALL_Z was added in 1.2.11. Change already made on 1.2 branch. Tue Jun 05 06:51:12 GMT 2012 Olly Betts <olly@survex.com> * common/compression_stream.cc,common/compression_stream.h: Move the default argument to the CompressionStream constructor to the header and mark the constructor as explicit, since it has a single argument form. Tue Jun 05 06:41:22 GMT 2012 Olly Betts <olly@survex.com> * api/queryvector.h: Fix somewhat dubious warning from clang. Tue Jun 05 06:33:09 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: Overhaul handling of compilers which pretend to be GCC, using a simple "case" on the preprocessed output of '__INTEL_COMPILER __clang__' to tell which is in use. We now explicitly check for clang, and only pass it warning flags it actually understands. GCC 4.0 is now lumped in with 3.*, since we handle it exactly the same. Enable -Wdouble-promotion for GCC >= 4.6. Check for symbol visibility support under any GCC-alike (which means we now run the test with Intel's compiler). Move the check for -Bsymbolic-functions to be run for all compilers. Mon Jun 04 06:21:02 GMT 2012 Olly Betts <olly@survex.com> * backends/dbfactory.cc: Test if we fail to open a stub database file in Xapian::Auto::open_stub() and throw an exception if so. * tests/api_backend.cc: Add regression test stubdb7. Sun Jun 03 11:43:44 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/query.h: Add second fake specialised form of the templated iterator Query constructor, for use in the Java bindings. Sat Jun 02 12:15:11 GMT 2012 Olly Betts <olly@survex.com> * docs/scalability.rst: Update gmane size, fix a typo, tweak wording. Sat Jun 02 12:10:55 GMT 2012 Olly Betts <olly@survex.com> * Makefile.am,api/omdocument.cc,api/omenquire.cc,backends/brass/, backends/chert/,bin/xapian-delve.cc,common/getopt.cc,configure.ac, docs/,m4-macros/xapian-1.3.m4,matcher/multimatch.cc,preautoreconf, tests/Makefile.am,tests/api_db.cc,tests/harness/testsuite.h, tests/harness/testutils.h: Change `...' quoting in prose to '...'. Fri Jun 01 07:40:48 GMT 2012 Olly Betts <olly@survex.com> * HACKING,Makefile.am: If you run "make coverage-check" by hand, the default of compressed HTML is unhelpful, so don't default to passing --html-gzip to genhtml, but instead add support for GENHTML_ARGS. Thu May 31 11:41:19 GMT 2012 Olly Betts <olly@survex.com> * tests/api_weight.cc: Add test that init() is now called for the term-independent weight contribution. Thu May 31 11:23:05 GMT 2012 Olly Betts <olly@survex.com> * weight/bm25weight.cc: Avoid use of undefined values when k1 or b are zero. Wed May 30 21:50:54 GMT 2012 Olly Betts <olly@survex.com> * tests/api_matchspy.cc: Use a cached generated database for matchspy2 and matchspy4. Wed May 30 13:22:54 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/weight.h,weight/weight.cc: We were failing to call init() for Weight objects providing the term-independent weight. These now get called with init(0.0). Wed May 30 13:13:03 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/weight.h: Remove duplicate "need_stat(WDF);" calls, accidentally added by merge of opsynonym branch in r12609. Wed May 30 07:44:23 GMT 2012 Olly Betts <olly@survex.com> * weight/bm25weight.cc: Fix use of uninitialised value found by new bm25weight4 testcase. Wed May 30 05:38:58 GMT 2012 Olly Betts <olly@survex.com> * tests/api_weight.cc: Add tests for BM25 parameter combinations where doclen shouldn't affect the weights. Wed May 30 05:11:34 GMT 2012 Olly Betts <olly@survex.com> * tests/Makefile.am: Move api_geospatial.cc into alphabetical order. Wed May 30 05:07:51 GMT 2012 Olly Betts <olly@survex.com> * weight/bm25weight.cc,weight/tradweight.cc: Throw SerialisationError not NetworkError if unserialise() fails. * tests/.gitignore,tests/Makefile.am,tests/api_weight.cc: Add test coverage for this. Wed May 30 05:01:33 GMT 2012 Olly Betts <olly@survex.com> * weight/tradweight.cc: BM25Weight -> TradWeight in exception message. Wed May 30 04:28:43 GMT 2012 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Fix feature test added for STEM_ALL_Z. Tue May 29 07:29:01 GMT 2012 Olly Betts <olly@survex.com> * tests/harness/testutils.cc,tests/harness/testutils.h: Remove unused test helper mset_range_is_same_percents() (thanks, lcov!) Tue May 29 04:10:07 GMT 2012 Olly Betts <olly@survex.com> * tests/generate-api_generated: Test that the string returned by a get_description() method isn't empty. Tue May 29 03:54:15 GMT 2012 Olly Betts <olly@survex.com> * generate-exceptions: Use function attributes in generated error.h. Tue May 29 02:10:23 GMT 2012 Olly Betts <olly@survex.com> * configure.ac,tests/runsrv.in: In the testsuite, only run remote servers under valgrind if x87 FP instructions are in use. Tue May 29 00:24:42 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/: Use the new function attributes on various API methods. * tests/api_wrdb.cc: get_termfreq() is now marked as pure, so the testsuite needs to actually use the return value to check its exception behaviour. Mon May 28 22:31:21 GMT 2012 Olly Betts <olly@survex.com> * api/query.cc,api/queryinternal.cc,api/queryinternal.h, include/xapian/query.h: Pass vector<pair<Xapian::termpos>, string> >& argument to Query::Internal::gather_terms() as a void* to avoid having to include <vector> in query.h. Mon May 28 05:07:02 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/attributes.h: Remove unused XAPIAN_NOTHROW_API_METHOD as we're handling that a different way instead. Mon May 28 00:41:55 GMT 2012 Olly Betts <olly@survex.com> * docs/doxygen_api.conf.in,docs/doxygen_source.conf.in: Hide new function attribute macros from doxygen. Sun May 27 14:04:34 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/query.h: Use "Internal" instead of "Query::Internal" in constructor prototype argument list for consistency with other cosntructors, and so this constructor gets ignored by SWIG. Sun May 27 07:38:56 GMT 2012 Olly Betts <olly@survex.com> * exception_data.pm: Fix typo to fix warning and so that XAPIAN_NOTHROW on a method will work. Sun May 27 01:43:30 GMT 2012 Olly Betts <olly@survex.com> * exception_data.pm,include/Makefile.mk,include/xapian.h, include/xapian/attributes.h,include/xapian/unicode.h: Add XAPIAN_CONST_FUNCTION, XAPIAN_PURE_FUNCTION, and XAPIAN_NOTHROW macros for marking functions and methods with those attributes. (tickets #151, #454) Fri May 25 06:42:35 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/weight.h: Fix documentation comment typo ("k1" -> "k"). Fri May 25 05:44:05 GMT 2012 Olly Betts <olly@survex.com> * backends/flint_lock.h: Mark FlintLock::throw_databaselockerror() as XAPIAN_NORETURN. Thu May 24 12:28:35 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/termgenerator.h,queryparser/termgenerator.cc, queryparser/termgenerator_internal.cc, queryparser/termgenerator_internal.h: Add TermGenerator::set_stemming_strategy() method, with strategies which correspond to those of QueryParser. Based on patch from Sehaj Singh Kalra, with some tweaks for adding term positions in more cases. (Fixes ticket#563) * tests/termgentest.cc: Add test coverage for the new features. Thu May 24 01:37:31 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Add doc comment for stem_strategy typedef. Wed May 23 11:33:08 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.lemony: Add new QueryParser::STEM_ALL_Z stemming strategy, which stems all terms and adds a Z prefix. (Patch from Sehaj Singh Kalra, fixes ticket#562) * tests/queryparsertest.cc: Add test coverage for QueryParser::STEM_ALL_Z. Wed May 23 05:37:20 GMT 2012 Olly Betts <olly@survex.com> * api/valuesetmatchdecider.cc,include/xapian/valuesetmatchdecider.h: "Limited" to "Ltd" in (C) two statements, for consistency (and with Charlie's agreement on IRC). Wed May 23 02:25:53 GMT 2012 Olly Betts <olly@survex.com> * api/,include/xapian/query.h: Rework QueryBranch to use a special QueryVector class rather than std::vector<Xapian::Query>. The special class is no bigger than std::vector<Xapian::Query>, and handles up to two entries by holding them within the object (at least GCC's std::vector doesn't currently do this space optimisation) which significantly reduces the memory used by a pairwise operator, which is very desirable as we no longer flatten a tree of the same pairwise operator as we build the query. Tue May 22 06:25:10 GMT 2012 Olly Betts <olly@survex.com> * api/queryinternal.cc: Handle the left side of AND_NOT and AND_MAYBE being MatchNothing in add_subquery() rather than in done(). Tue May 22 02:40:45 GMT 2012 Olly Betts <olly@survex.com> * api/queryinternal.cc: Handle QueryAndLike with a MatchNothing subquery in add_subquery() rather than done(). Tue May 22 02:01:00 GMT 2012 Olly Betts <olly@survex.com> * api/queryinternal.cc,api/queryinternal.h: Drop MatchNothing subqueries in OR-like situations in add_subquery() rather than adding them and then handling it later. Sat May 19 00:27:23 GMT 2012 Olly Betts <olly@survex.com> * tests/runsrv.in: Add explanation of why we have this script. Sat May 19 00:27:10 GMT 2012 Olly Betts <olly@survex.com> * tests/runtest.in: Fix comment typo. Fri May 18 12:50:08 GMT 2012 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc: Check if the database is closed when asked to iterate metadata keys. Fixes recently added testcase closedb10 for inmemory. Fri May 18 12:34:24 GMT 2012 Olly Betts <olly@survex.com> * tests/Makefile.am: Pass -ldl last when compiling zlib-vg.so, as that seems to be needed on Ubuntu 12.04. Fri May 18 04:54:09 GMT 2012 Olly Betts <olly@survex.com> * tests/api_closedb.cc: We now test almost all methods of Database and WritableDatabase after calling close(). (ticket#337) Thu May 17 14:26:11 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/,backends/chert/: After closing the database, methods which try to use the termlist would throw FeatureUnavailableError with message "Database has no termlist", assuming that the termlist table not being open meant it wasn't present. Fix to check if the postlist_table is open to determine which case we're in. * tests/api_closedb.cc: Improve test coverage for closed databases. (ticket#337) Tue May 15 11:24:34 GMT 2012 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog and 1.2.10. Tue May 15 11:12:42 GMT 2012 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.cc: Fix incorrect use of "delete" to "delete []". The type is POD, so I suspect this mistake didn't actually cause any problems, and it only happens when a cursor gets rebuilt because the B-tree has gained a level, so it's a rather rare occurrence. Tue May 15 11:09:44 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_cursor.cc: Fix incorrect use of "delete" to "delete []". The type is POD, so I suspect this mistake didn't actually cause any problems, and it only happens when a cursor gets rebuilt because the B-tree has gained a level, so it's a rather rare occurrence. Thu May 10 02:16:21 GMT 2012 Olly Betts <olly@survex.com> * HACKING: Update Debian/Ubuntu packaging instructions. Thu May 10 02:14:48 GMT 2012 Olly Betts <olly@survex.com> * HACKING: freshmeat -> freecode. Wed May 09 02:54:40 GMT 2012 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Add "safeunistd.h", required for GCC 4.7. Wed May 09 00:09:17 GMT 2012 Olly Betts <olly@survex.com> * api/query.cc: Construct MatchAll using constructor rather than assignment syntax. Wed May 09 00:08:29 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_cursor.h: Whitespace tweaks. Tue May 08 23:51:01 GMT 2012 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Document xapian-check for fixing corrupted databases. Tue May 08 11:30:06 GMT 2012 Olly Betts <olly@survex.com> * common/fileutils.cc: Add safeunistd.h for mkdir, required by GCC 4.7 (reported by Gaurav Arora). Wed May 02 03:44:25 GMT 2012 Olly Betts <olly@survex.com> * weight/weightinternal.cc: Using AssertEq() on NULL doesn't compile, at least with recent GCC. Wed May 02 02:58:37 GMT 2012 Olly Betts <olly@survex.com> * common/omassert.cc: Fix warning with GCC in build with assertions enabled. Wed May 02 02:55:54 GMT 2012 Olly Betts <olly@survex.com> * HACKING: We're now using automake 1.12 to generate snapshots and releases. Mon Apr 23 13:16:36 GMT 2012 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Fix tradweight1 to test that TradWeight(0) means that wdf and doc length don't affect the weight of a term. Mon Apr 23 03:44:02 GMT 2012 Olly Betts <olly@survex.com> * backends/chert/,backends/dbcheck.cc,bin/xapian-check.cc, include/xapian/database.h: Add "fix" option to xapian-check, which currently will regenerate iamchert if it isn't valid, and will regenerate base files from the .DB files (only really tested on databases which have just been compacted). Mon Apr 23 03:40:35 GMT 2012 Olly Betts <olly@survex.com> * common/filetests.h: Add new function file_size() to get the size of a file from a path or file descriptor. * backends/brass/brass_compact.cc,backends/chert/chert_compact.cc, net/remoteconnection.cc,tests/api_replicate.cc: Use this new function. Wed Apr 18 01:35:21 GMT 2012 Olly Betts <olly@survex.com> * weight/bm25weight.cc,weight/tradweight.cc: Fix comment typo. Wed Apr 11 08:26:42 GMT 2012 Olly Betts <olly@survex.com> * HACKING: Update section on patches to mention git (git diff and git format-patch), and using "-r" with normal diff, and also that ptardiff offers a nice way to diff against an unpacked tarball. Wed Apr 04 02:22:04 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc,backends/chert/chert_compact.cc: If the output database is empty, don't write a metainfo tag out. Take care not to divide by zero when computing the percentage size change for a table. * tests/api_compact.cc: Add new testcase compactempty1 to provide regression tests. Wed Apr 04 01:48:34 GMT 2012 Olly Betts <olly@survex.com> * backends/dbcheck.cc: If the database couldn't be opened, don't try to reserve space for cross-checking doclens, as that just results in us reporting a confusing warning about there being too many documents to cross-check doclens. Fri Mar 30 22:23:53 GMT 2012 Olly Betts <olly@survex.com> * backends/document.h: "An string" -> "A string" in internal doc comment. Fri Mar 30 22:20:11 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/weight.h: Fix switched lower and upper in doc comments for Weight methods get_doclength_lower_bound() and get_doclength_upper_bound(). Correct maximum to minimum in get_doclength_lower_bound() comment and note that this excludes zero length documents. Fix "An lower" to "A lower". Wed Mar 28 09:54:33 GMT 2012 Olly Betts <olly@survex.com> * NEWS: Fix typos in another old entry. Wed Mar 28 09:53:00 GMT 2012 Olly Betts <olly@survex.com> * NEWS: DatabaseCoruptError -> DatabaseCorruptError in old entries. Wed Mar 28 08:05:13 GMT 2012 Olly Betts <olly@survex.com> * tests/termgentest.cc: Add test coverage for discarding of terms > 64 bytes by TermGenerator. Tue Mar 27 21:07:07 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/database.h: Note when Database::close() was added. Tue Mar 27 13:03:22 GMT 2012 Olly Betts <olly@survex.com> * api/omenquire.cc,docs/deprecation.rst,include/xapian/enquire.h, include/xapian/errorhandler.h: Deprecate Xapian::ErrorHandler. (ticket#3) Tue Mar 27 13:01:53 GMT 2012 Olly Betts <olly@survex.com> * HACKING,docs/deprecation.rst,docs/doxygen_api.conf.in, docs/doxygen_source.conf.in: Updates for addition of XAPIAN_DEPRECATED_EX() and XAPIAN_DEPRECATED_CLASS_EX. Tue Mar 27 10:11:23 GMT 2012 Olly Betts <olly@survex.com> * HACKING,include/xapian.h,include/xapian/deprecated.h: Add XAPIAN_DEPRECATED_EX() and XAPIAN_DEPRECATED_CLASS_EX macros which mark a feature as deprecated externally but not when building the library. Tue Mar 27 08:46:29 GMT 2012 Olly Betts <olly@survex.com> * api/compactor.cc: Include the xapian sub-headers we actually use rather than xapian.h, to avoid needless rebuilds. Tue Mar 27 08:33:28 GMT 2012 Olly Betts <olly@survex.com> * HACKING: Add example of deprecating a pure virtual method. Tue Mar 27 08:30:17 GMT 2012 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Sat Mar 24 12:33:44 GMT 2012 Olly Betts <olly@survex.com> * backends/brass/brass_databasereplicator.cc: Fix signed vs unsigned comparison warning with GCC 4.4.3. Sat Mar 24 12:26:13 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser_internal.h: Change the default stemming strategy to STEM_SOME, to eliminate the API gotcha that setting a stemmer is ignored until you also set a strategy. * tests/queryparsertest.cc: Add testcase qp_defaultstrategysome1 to check that the default is STEM_SOME. Fri Mar 23 04:29:32 GMT 2012 Olly Betts <olly@survex.com> * tests/perftest/perftest.cc: Fix comment cut-and-paste error. Fri Mar 23 04:28:24 GMT 2012 Olly Betts <olly@survex.com> * Makefile.am: Use git commit hash in the title of a coverage report generated from a git tree. Thu Mar 22 04:29:38 GMT 2012 Olly Betts <olly@survex.com> * xapian-core.spec.in: Remove xapian-chert-update. Mon Mar 19 05:42:50 GMT 2012 Olly Betts <olly@survex.com> * matcher/multiandpostlist.cc: Don't count unweighted subqueries of MultiAndPostList for percentage calculations, as OP_FILTER maps to MultiAndPostList now. (ticket#590) * tests/api_percentages.cc: Regression test topercent6. Sun Mar 18 23:24:04 GMT 2012 Dan Colish <dcolish@> * backends/brass/brass_databasereplicator.cc: size out buffer to the incoming blocksize Sun Mar 18 17:02:14 GMT 2012 Dan Colish <dcolish@> * .gitignore: ignore /.deps as well Sun Mar 18 16:46:45 GMT 2012 Dan Colish <dcolish@> * .gitignore:Update gitignore for bump to libxapian-1.3 Sun Mar 18 04:37:37 GMT 2012 Olly Betts <olly@survex.com> * configure.ac: Set LIBRARY_VERSION_SUFFIX to -1.3. Sun Mar 18 00:48:05 GMT 2012 Dan Colish <dcolish@gmail.com> * backends/brass/,common/Makefile.mk, common/compression_stream.h, common/compression_stream.cc: Compress changesets in brass replication. Increments the changeset version. Ticket #348 Sun Mar 18 00:39:23 GMT 2012 Dan Colish <dcolish@gmail.com> * tests/.gitignore,tests/soaktest/.gitignore: Add additional ignores for test artifacts Fri Mar 16 05:23:54 GMT 2012 Olly Betts <olly@survex.com> * common/Makefile.mk: Missing part of previous change. Fri Mar 16 04:59:53 GMT 2012 Olly Betts <olly@survex.com> * Makefile.am,m4-macros/xapian.m4: Rename libxapian to libxapian-1.3 and xapian.m4 to xapian-1.3.m4. Wed Mar 14 04:58:55 GMT 2012 Olly Betts <olly@survex.com> * NEWS,tests/api_unicode.cc,unicode/tclUniData.cc: Update Unicode character database to 6.1.0. Wed Mar 14 02:23:27 GMT 2012 Olly Betts <olly@survex.com> * NEWS: Update for 1.3.0. Tue Mar 13 07:44:32 GMT 2012 Olly Betts <olly@survex.com> * NEWS: Update from 1.2.9 and ChangeLog. Tue Mar 13 00:28:58 GMT 2012 Olly Betts <olly@survex.com> * docs/geospatial.rst: Fix typo - it's LatLongCoords::append() not LatLongCoords::insert(). Mon Mar 12 23:14:20 GMT 2012 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Fix typo - need "1.2.x" not "1.0.x" to get both flint and chert support. Mon Mar 12 11:54:19 GMT 2012 Olly Betts <olly@survex.com> * docs/Makefile.am: doxygen no longer generates any GIF files, so stop trying to ship or install them. Thu Mar 08 10:21:51 GMT 2012 Olly Betts <olly@survex.com> * docs/index.rst: Add link to new "getting started" guide. Mon Mar 05 06:05:28 GMT 2012 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Fix curly double quote testcase. Mon Mar 05 00:36:33 GMT 2012 Olly Betts <olly@survex.com> * matcher/: Fix issue with AND, OR, and XOR queries against a database with no documents in it - this was causing a divide by zero, which led to MSet::get_matches_estimated() reporting 2147483648 on i386. * tests/api_backend.cc: Add regression test emptydb1. Sun Mar 04 23:05:17 GMT 2012 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix comment typo. Fri Mar 02 04:00:19 GMT 2012 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Address new warnings from GCC 4.6. Wed Feb 29 22:36:26 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/database.h: Improve documentation of Database::close() and ~WritableDatabase(). Thu Jan 26 03:49:24 GMT 2012 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Allow Unicode curly double quote characters to start and/or end phrases. Sun Jan 22 13:28:30 GMT 2012 Olly Betts <olly@survex.com> * bin/xapian-delve.cc: Send errors to stderr not stdout. Sun Jan 22 13:16:21 GMT 2012 Olly Betts <olly@survex.com> * HACKING: Update details of versions of doxygen, automake and libtool used to bootstrap snapshots and releases. Fri Jan 20 21:35:38 GMT 2012 Olly Betts <olly@survex.com> * tests/Makefile.am: unittest needs getopt for non-glibc platforms. Fri Jan 20 21:33:59 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/query.h: GCC 4.2.1 on Mac OS X seems to need a copy ctor for InvertedQuery_ - unclear why, but it's easy to add one. Wed Jan 18 13:13:07 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/matchspy.h,include/xapian/postingsource.h, include/xapian/weight.h: Document that you can define a static operator delete method in your subclass if deallocation needs to be handled specially. (Closes ticket#554) Wed Jan 18 12:27:20 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/geospatial.h: Mark as experimental for now. Wed Jan 18 12:22:39 GMT 2012 Olly Betts <olly@survex.com> * docs/geospatial.rst: Say "haversine" rather than "Haversine" consistently. Wed Jan 18 12:01:28 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/geospatial.h: Change empty() to return bool not size_t. Wed Jan 18 10:44:16 GMT 2012 Olly Betts <olly@survex.com> * docs/geospatial.rst: Fix minor typos in geospatial docs. Wed Jan 18 10:39:53 GMT 2012 Olly Betts <olly@survex.com> * Makefile.am,api/postingsource.cc,api/queryinternal.cc, api/registry.cc,common/output.h,common/registryinternal.h, docs/geospatial.rst,geospatial/,include/Makefile.mk,include/xapian.h, include/xapian/geospatial.h,include/xapian/postingsource.h, include/xapian/registry.h,tests/.gitignore,tests/Makefile.am, tests/api_geospatial.cc: Merge geotomerge2 branch from github. Closes ticket#481. Tue Jan 17 22:29:10 GMT 2012 Olly Betts <olly@survex.com> * api/errorhandler.cc: Reorder header includes. Mon Jan 16 22:59:02 GMT 2012 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.lemony: Fix FLAG_AUTO_SYNONYMS not to enable auto multi-word synonyms too. * tests/queryparsertest.cc: Extend testcase qp_synonym1 to include a regression test for this bug. Sun Jan 15 12:41:35 GMT 2012 Olly Betts <olly@survex.com> * .gitignore,backends/Makefile.mk,backends/brass/,backends/chert/, backends/dbcheck.cc,bin/,common/bitstream.h, include/xapian/database.h: Add a (currently experimental) API for checking the integrity of databases (partly addresses ticket#238). This means we can remove XAPIAN_VISIBILITY_DEFAULT from a number of functions which aren't in the public API (partly addresses ticket#63). Fri Jan 13 12:07:39 GMT 2012 Olly Betts <olly@survex.com> * tests/stemtest.cc: Include <xapian.h> rather than <xapian/stem.h> - we're trying to test the public API, so including the public API header is better. Thu Jan 12 13:15:27 GMT 2012 Olly Betts <olly@survex.com> * tests/unittest.cc: Add new testcases simple_exceptions_work1 and class_exceptions_work1 which perform sanity tests of exception handling. * tests/internaltest.cc: Remove testcase test_except1 which is now redundant. Thu Jan 12 13:09:25 GMT 2012 Olly Betts <olly@survex.com> * common/str.cc: Special cases for 0 in tostring() templates aren't actually needed, but special casing single digit numbers may be worthwhile anyway, so change comments to FIXME ones about checking this. * tests/internaltest.cc: Test str() on unsigned 0, and some negative, signed positive, and unsigned positive single digits values. Wed Jan 11 13:03:54 GMT 2012 Olly Betts <olly@survex.com> * backends/multi/multi_alltermslist.h: Fix filename in @file. Wed Jan 11 12:53:44 GMT 2012 Olly Betts <olly@survex.com> * api/documentterm.h,api/maptermlist.h,api/omenquireinternal.h, backends/brass/,backends/chert/,backends/database.h, backends/document.h,backends/inmemory/inmemory_alltermslist.h, backends/inmemory/inmemory_database.h, backends/inmemory/inmemory_positionlist.h, backends/multi/multi_postlist.h,backends/multi/multi_termlist.h, backends/positionlist.h,backends/remote/net_termlist.h, backends/valuestats.h,common/,include/xapian/query.h,matcher/, queryparser/queryparser_internal.h,tests/harness/, tests/perftest/freemem.h,tests/perftest/runprocess.h: Convert comments at the start of files to @file and @brief doxygen comments. Wed Jan 11 11:36:55 GMT 2012 Olly Betts <olly@survex.com> * matcher/phrasepostlist.h: Remove lone "----START-LICENCE----" - the rest got removed ages ago. Wed Jan 11 11:23:04 GMT 2012 Olly Betts <olly@survex.com> * common/replicate_utils.h: Fix "@raises" to "@exception". Wed Jan 11 11:21:31 GMT 2012 Olly Betts <olly@survex.com> * matcher/multiandpostlist.h: Fix "@parameter" to "@param". Wed Jan 11 11:20:11 GMT 2012 Olly Betts <olly@survex.com> * common/debuglog.h: Fix "@msg" to "@a msg" in doxygen comment. Wed Jan 11 11:18:50 GMT 2012 Olly Betts <olly@survex.com> * matcher/multiandpostlist.h: Correct @exceptions to @exception. Wed Jan 11 10:48:05 GMT 2012 Olly Betts <olly@survex.com> * HACKING: We no longer have any uses of \ to introduce doxygen directives, so update HACKING to reflect that. Wed Jan 11 10:46:01 GMT 2012 Olly Betts <olly@survex.com> * api/omdatabase.cc: Change \todo to @todo. Wed Jan 11 10:09:29 GMT 2012 Olly Betts <olly@survex.com> * api/replication.h,api/valuesetmatchdecider.cc, common/output-internal.h,common/output.h,include/xapian/, tests/harness/testrunner.cc,tests/harness/testrunner.h, tests/perftest/perftest.cc,tests/perftest/perftest.h: Change the instances of \file and \brief in doxygen comments to @file and @brief. Wed Jan 11 10:08:19 GMT 2012 Olly Betts <olly@survex.com> * tests/internaltest.cc: Remove serialisedoc1 from the list of testcases to run too. Wed Jan 11 09:50:22 GMT 2012 Olly Betts <olly@survex.com> * matcher/multixorpostlist.h: Remove "using namespace std;" which we don't need here. Wed Jan 11 09:38:27 GMT 2012 Olly Betts <olly@survex.com> * common/filetests.h: Remove "using namespace std;" which we don't need here. Wed Jan 11 09:36:27 GMT 2012 Olly Betts <olly@survex.com> * common/filetests.h: Add documentation comments. Wed Jan 11 09:28:26 GMT 2012 Olly Betts <olly@survex.com> * tests/api_serialise.cc: Extend serialise_document1 to test serialising and unserialising an empty document, which is an interesting edge case. * tests/internaltest.cc: Remove testcase serialisedoc1, which now doesn't add useful additional coverage over serialise_document1. Wed Jan 11 08:29:04 GMT 2012 Olly Betts <olly@survex.com> * tests/internaltest.cc: Test document serialisation via the public API (which was added after the tests were). * net/serialise.h: Remove XAPIAN_VISIBILITY_DEFAULT from serialise_document() and unserialise_document(). Wed Jan 11 05:57:14 GMT 2012 Olly Betts <olly@survex.com> * api/replication.cc,common/,tests/unittest.cc: Move removedir() from utils.{cc,h} to fileutils.{cc,h}, and remove the now empty utils.{cc,h}. Wed Jan 11 05:44:38 GMT 2012 Olly Betts <olly@survex.com> * tests/harness/unixcmds.cc: Check return values from system() calls and throw a std::string exception if non-zero. Wed Jan 11 05:20:30 GMT 2012 Olly Betts <olly@survex.com> * common/Makefile.mk,common/omassert.cc,common/omassert.h, common/utils.cc: Change the sense of the bool returned by within_DBL_EPSILON() to match what the name suggests instead of being the opposite, and change the use of it too. Move within_DBL_EPSILON() implementation from utils.cc to new file omassert.cc, only compile it when assertions are on, and prototype it in omassert.h rather than utils.h. Wed Jan 11 05:09:39 GMT 2012 Olly Betts <olly@survex.com> * api/compactor.cc,api/replication.cc,backends/brass/,backends/chert/, backends/dbfactory.cc,common/utils.cc,common/utils.h, tests/api_backend.cc,tests/api_wrdb.cc, tests/harness/backendmanager.cc,tests/harness/unixcmds.cc: Remove std::string wrappers for C library functions from utils.h and just explicitly call c_str() when we need to. These wrappers are problematic as they suppress warnings from _FORTIFY_SOURCE about ignored return values, and I can't see an easy way to avoid that. Wed Jan 11 03:35:35 GMT 2012 Olly Betts <olly@survex.com> * api/replication.cc,backends/brass/brass_databasereplicator.cc, backends/brass/brass_table.cc, backends/chert/chert_databasereplicator.cc, backends/chert/chert_table.cc,backends/databasereplicator.cc, backends/dbfactory.cc,common/utils.cc,common/utils.h: Use file_exists() and dir_exists() from common/filetests.h everywhere and remove the old versions in common/utils.h and common/utils.cc. Wed Jan 11 02:20:45 GMT 2012 Olly Betts <olly@survex.com> * common/utils.h: Remove XAPIAN_VISIBILITY_DEFAULT from old file_exists() and dir_exists(). Wed Jan 11 02:19:58 GMT 2012 Olly Betts <olly@survex.com> * tests/api_compact.cc: Convert to use filetests.h. Wed Jan 11 02:14:07 GMT 2012 Olly Betts <olly@survex.com> * bin/xapian-check.cc,bin/xapian-inspect.cc: Convert to use filetests.h. Wed Jan 11 02:13:02 GMT 2012 Olly Betts <olly@survex.com> * common/filetests.h: Rename directory_exists() to dir_exists(). Wed Jan 11 02:10:06 GMT 2012 Olly Betts <olly@survex.com> * common/Makefile.mk,common/filetests.h,tests/api_replicate.cc, tests/harness/backendmanager_multi.cc,tests/harness/testsuite.cc: Add new inline versions of file_exists() and directory_exists() and use these from the testsuite, as a step towards making these internal to the library. Tue Jan 10 23:37:36 GMT 2012 Olly Betts <olly@survex.com> * tests/api_serialise.cc: Change the rather odd "operator delete(p);" to the more natural "delete p;". Tue Jan 10 22:45:01 GMT 2012 Olly Betts <olly@survex.com> * tests/unittest.cc: Fix message typo in recent unittest change. Tue Jan 10 13:48:38 GMT 2012 Olly Betts <olly@survex.com> * common/safesysstat.h: The S_ISDIR and S_ISREG macros for MSVC are unlikely to work on other platforms (_S_IFMT vs S_IFMT, etc) so split them. Tweak a few comments. Tue Jan 10 13:10:25 GMT 2012 Olly Betts <olly@survex.com> * net/: Split encode_length() and decode_length() out of serialise.{cc,h} into new file length.{cc,h}, and remove XAPIAN_VISIBILITY_DEFAULT from decode_length(). * api/,backends/brass/brass_database.cc, backends/chert/chert_database.cc,backends/remote/net_postlist.cc, backends/remote/remote-database.cc,net/: Update headers which need to be included. * tests/internaltest.cc,tests/unittest.cc: Move serialiselength1 and serialiselength2 from internaltest to unittest. Mon Jan 09 02:16:22 GMT 2012 Olly Betts <olly@survex.com> * tests/api_compact.cc: Actually add test coverage for ByteLengthPrefixedStringItor - the previous attempt didn't cause merging of synonym lists to happen. Mon Jan 09 00:21:25 GMT 2012 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Fix comment typo. Sun Jan 08 14:47:19 GMT 2012 Olly Betts <olly@survex.com> * tests/api_compact.cc: Add test coverage for ByteLengthPrefixedStringItor. Sun Jan 08 14:05:19 GMT 2012 Olly Betts <olly@survex.com> * common/fd.h,tests/api_db.cc,tests/harness/backendmanager.cc: Add missing explicit header includes, uncovered by compiling with STLport. Sat Jan 07 11:45:00 GMT 2012 Olly Betts <olly@survex.com> * INSTALL,configure.ac,m4/rjb_find_stlport.m4,xapian-config.in: Remove configure's --with-stlport and --with-stlport-compiler options, as they don't allow you to actually specify what you need (at least to use the Debian STLport package), and instead document what to pass to configure to enable building with STLport (though it seems to no longer be actively maintained, and the debug mode (which is probably the most interesting feature now) doesn't seem to work on Debian stable). Sat Jan 07 10:24:25 GMT 2012 Olly Betts <olly@survex.com> * HACKING: Update STLPort URL to point to the more active fork on SourceForge. Fri Jan 06 20:07:06 GMT 2012 Olly Betts <olly@survex.com> * tests/api_sorting.cc: Add check that NeverUseMeKeyMaker::operator() would actually cause a test failure if called. Fri Jan 06 20:01:27 GMT 2012 Olly Betts <olly@survex.com> * tests/api_stem.cc: Test get_description() on a user-implemented stemmer. Fri Jan 06 14:34:16 GMT 2012 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/harness/,tests/queryparsertest.cc, tests/termgentest.cc,tests/unittest.cc: Make unittest use the test harness, so it gets all the valgrind and fd leak checks, and other handy features all the other tests have. Currently this is done in a bit of a hacky way, but it's a starting point for abstracting out the Xapian-specific parts of the test harness. Fri Jan 06 08:42:26 GMT 2012 Olly Betts <olly@survex.com> * Makefile.am: Delete generated lcov directory before running genhtml. Pass --demangle-cpp and --html-gzip to genhtml. Fri Jan 06 08:14:05 GMT 2012 Olly Betts <olly@survex.com> * Makefile.am: Trim `pwd` off all paths in coverage report. Fri Jan 06 02:29:04 GMT 2012 Olly Betts <olly@survex.com> * common/pretty.h: Add support for AndContext and OrContext. Fri Jan 06 02:28:07 GMT 2012 Olly Betts <olly@survex.com> * common/pretty.h: Use XAPIAN_PRETTY_AS_CLASSNAME for classes in namespace Xapian::Internal too. Thu Jan 05 12:23:10 GMT 2012 Olly Betts <olly@survex.com> * tests/internaltest.cc,tests/unittest.cc: Move tests of serialise_double() and unserialise_double() from internaltest to unittest. * common/serialise-double.h: This means serialise_double() and unserialise_double() no longer need external visibility, so make them internal only. Thu Jan 05 12:08:15 GMT 2012 Olly Betts <olly@survex.com> * common/bitstream.h: BitWriter doesn't need public visibility. Thu Jan 05 10:53:23 GMT 2012 Olly Betts <olly@survex.com> * common/output-internal.h,common/pretty.h: Fix include paths for two cases which are only used in a logging build. Fri Dec 30 09:17:50 GMT 2011 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Update expected results (mostly because we no longer flatten groups of the same operator at Query construction time). Thu Dec 29 13:47:04 GMT 2011 Olly Betts <olly@survex.com> * api/Makefile.mk: Add missing backslash so various headers actually get shipped. Thu Dec 29 11:29:04 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/query.h: Add fake templated ctor for SWIG. Thu Dec 29 00:23:25 GMT 2011 Olly Betts <olly@survex.com> * tests/api_scalability.cc: Add regression test querypairwise1 for ticket#273. Thu Dec 29 00:03:17 GMT 2011 Olly Betts <olly@survex.com> * api/,backends/remote/remote-database.cc, backends/remote/remote-database.h,docs/deprecation.rst, include/xapian/query.h,matcher/,net/remoteserver.cc, queryparser/queryparser.lemony,tests/,weight/weightinternal.h: Merge query-internal-reimplementation git branch which reimplements Query::Internal. (ticket#280). These are the most notable changes: + Query objects are smaller and should be faster. + More readable format for Query::get_description(). + More compact serialisation format for Query objects. + Query operators are no longer flattened as you build up a tree (but the query optimiser still combines groups of the same operator). This means that Query objects are truly immutable, and so we don't need to copy Query objects when composing them. This should also fix a few O(n*n) cases when building up an n-way query pair-wise. (ticket#273) + The Query optimiser can do a few extra optimisations. Mon Dec 26 12:13:36 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.cc: Change QueryParser::set_default_op() to reject operators which don't make sense to set, and explicitly document all the operators which are allowed. * tests/queryparsertest.cc: Check that bad operators are rejected in existing testcase qp_default_op2. Add new testcase qp_default_op3 to check that good operators can be set and work in a simple testcase. Mon Dec 26 09:12:09 GMT 2011 Olly Betts <olly@survex.com> * tests/harness/backendmanager_brass.h, tests/harness/backendmanager_chert.h: Remove superfluous "private:" access specifiers. Sat Dec 24 12:34:19 GMT 2011 Olly Betts <olly@survex.com> * HACKING: Recommend installing valgrind, ccache, and eatmydata. Sat Dec 24 12:28:44 GMT 2011 Olly Betts <olly@survex.com> * HACKING: Make it clear "Building from SVN" applies to git too, and mention that git-svn is needed when building from git. Thu Dec 22 11:18:17 GMT 2011 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Initialise variable which can be used uninitialised. Thu Dec 22 03:14:02 GMT 2011 Olly Betts <olly@survex.com> * bin/Makefile.mk: Link $(libxapian_la) after libbrasscheck.la and libchertcheck.la to fix mingw build. Probably closes ticket#567. Thu Dec 22 02:45:45 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/query.h: Document that OP_ELITE_SET with non-term subqueries might pick subqueries which don't match anything. Closes ticket#49. Thu Dec 22 02:37:10 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/query.h: Improve documentation of OP_ELITE_SET. Wed Dec 21 12:26:00 GMT 2011 Olly Betts <olly@survex.com> * api/errorhandler.cc: Rearrange code in ErrorHandler::operator() to be clearer. Wed Dec 21 11:23:28 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/errorhandler.h: Fix typo: 'APU' -> 'API'. Wed Dec 21 09:57:32 GMT 2011 Olly Betts <olly@survex.com> * api/,backends/brass/brass_postlist.cc, backends/brass/brass_postlist.h,backends/chert/, backends/contiguousalldocspostlist.cc, backends/contiguousalldocspostlist.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/multi/multi_postlist.cc,backends/multi/multi_postlist.h, backends/remote/,docs/,expand/,include/xapian/,matcher/, net/remoteserver.cc,net/serialise.cc,tests/,weight/bm25weight.cc, weight/boolweight.cc,weight/tradweight.cc: Deprecate Xapian::weight typedef in favour of just using double, and make the replacement everywhere. Closes ticket#560. Tue Dec 20 13:31:21 GMT 2011 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Include "safesysstat.h" for fstat(). Fri Dec 16 03:05:17 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_termlisttable.cc, backends/chert/chert_termlisttable.cc, backends/multi/multi_postlist.cc,net/replicatetcpclient.cc, net/serialise.cc,net/tcpserver.cc,tests/, tests/harness/backendmanager_remoteprog.cc,tests/queryparsertest.cc, tests/soaktest/soaktest_queries.cc,tests/termgentest.cc: Remove unnecessary includes on "utils.h". Fri Dec 16 02:57:49 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_btreebase.cc,backends/brass/brass_database.cc, backends/brass/brass_databasereplicator.cc, backends/chert/chert_btreebase.cc,backends/chert/chert_database.cc, backends/chert/chert_databasereplicator.cc, net/remoteconnection.cc,tests/api_replicate.cc: Replace uses of fdcloser class with new FD class. * common/utils.h: Remove fdcloser class. Fri Dec 16 00:56:45 GMT 2011 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Use a const int rather than a #define for BUFSIZE. Fri Dec 16 00:50:57 GMT 2011 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Avoid leaking fds on exceptions in truncated_copy(). Fri Dec 16 00:42:11 GMT 2011 Olly Betts <olly@survex.com> * common/fd.h: Oops, fix to actually compile. Thu Dec 15 23:54:44 GMT 2011 Olly Betts <olly@survex.com> * common/Makefile.mk,common/fd.h: Add new FD class which wraps a file descriptor and ensures it gets released when we exit the scope. Thu Dec 15 22:44:16 GMT 2011 Olly Betts <olly@survex.com> * weight/Makefile: Add forwarding Makefile. Thu Dec 15 22:17:18 GMT 2011 Olly Betts <olly@survex.com> * weight/Makefile.mk: Ship dir_contents and Makefile. Tue Dec 13 10:33:21 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog and branches/1.2. Fri Dec 09 10:33:25 GMT 2011 Olly Betts <olly@survex.com> * tests/soaktest/soaktest.cc,tests/soaktest/soaktest.h, tests/soaktest/soaktest_queries.cc: "boulton" -> "Boulton" in copyright statements. Fri Dec 09 10:31:18 GMT 2011 Olly Betts <olly@survex.com> * tests/soaktest/soaktest.cc: With Sun's compiler, random() and srandom() aren't in <cstdlib> so we need to use <stdlib.h> instead. Fri Dec 09 07:51:44 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/types.h: Set BAD_VALUENO to 0xffffffff rather than static_cast<valueno>(-1) - it's the same value currently, but more robust to valueno being set a wider type. Fri Dec 09 07:49:07 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/types.h: Explicitly document which value slot numbers are valid. Fixes ticket#555. Sat Dec 03 09:32:55 GMT 2011 Olly Betts <olly@survex.com> * tests/api_opsynonym.cc: Fix typo in comment. Thu Dec 01 04:41:24 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/database.h: Remove bogus paragraph in replace_document() documentation comment, cut and pasted from delete_document() documentation comment. (Fixes bug#579) Tue Nov 08 01:50:25 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Add missing doxygen @param commands. Tue Nov 08 01:33:34 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/database.h,include/xapian/document.h: Add missing doxygen @param commands. Mon Nov 07 10:31:04 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Address doxygen warnings. Mon Nov 07 02:40:50 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/postingsource.h: Document undocumented parameters, fixing doxygen warnings. * include/xapian/queryparser.h: Escape < and > in doxygen comments. Mon Nov 07 02:38:18 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Enquire::get_eset() doesn't take a min_wt parameter, so remove @param documentation for it. Mon Nov 07 01:59:54 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/: Document lots of parameters, fixing doxygen warnings. Sun Nov 06 22:05:09 GMT 2011 Olly Betts <olly@survex.com> * generate-exceptions: Document all parameters to fix doxygen warnings. Sun Nov 06 22:03:50 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/: Fix doxygen warnings. Sun Nov 06 21:00:08 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Document value parameter of sortable_serialise() and sortable_unserialise(). Sun Nov 06 17:53:31 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/document.h: Improve Document get_data() and set_data() documentation comments. Sun Nov 06 17:14:39 GMT 2011 Olly Betts <olly@survex.com> * tests/Makefile.am: Add -I$(top_srcdir) so api/replication.h is found in a VPATH build. Sun Nov 06 16:06:52 GMT 2011 Olly Betts <olly@survex.com> * api/,backends/database.h,backends/document.h, backends/multi/multi_alltermslist.h,backends/positionlist.h,common/, tests/harness/: More more headers out of common. Sun Nov 06 14:55:25 GMT 2011 Olly Betts <olly@survex.com> * api/maptermlist.h,backends/inmemory/Makefile.mk, backends/inmemory/inmemory_positionlist.h, backends/remote/Makefile.mk,backends/remote/remote-database.cc, backends/remote/remote-database.h,common/Makefile.mk, common/inmemory_positionlist.h,common/remote-database.h, matcher/multimatch.cc,matcher/remotesubmatch.cc, matcher/remotesubmatch.h,net/progclient.h,net/remotetcpclient.h: Move two more headers out of common. Sun Nov 06 11:56:46 GMT 2011 Olly Betts <olly@survex.com> * tests/internaltest.cc: Fix for VPATH build. Sun Nov 06 11:26:18 GMT 2011 Olly Betts <olly@survex.com> * common/,tests/api_replicate.cc: Factor internal classes out of output.h into new output-internal.h to fix build failures in VPATH due to move of weightinternal.h out of common (since the test harness uses output.h). Sun Nov 06 00:49:45 GMT 2011 Olly Betts <olly@survex.com> * api/,backends/brass/,backends/chert/,backends/dbfactory_remote.cc, backends/inmemory/inmemory_database.cc, backends/multi/multi_termlist.cc,backends/remote/net_postlist.cc, backends/remote/net_termlist.h,backends/remote/remote-database.cc, bin/,common/,expand/,matcher/,net/,tests/internaltest.cc, weight/Makefile.mk,weight/weightinternal.h: Move even more headers out of common and next to their corresponding C++ file. Sat Nov 05 23:55:55 GMT 2011 Olly Betts <olly@survex.com> * api/replication.cc,backends/Makefile.mk, backends/brass/brass_databasereplicator.h, backends/chert/chert_databasereplicator.h, backends/databasereplicator.h,common/Makefile.mk, common/databasereplicator.h: Move databasereplicator.h from common to backends. Sat Nov 05 23:51:40 GMT 2011 Olly Betts <olly@survex.com> * backends/Makefile.mk: Fix sort order of noinst_HEADERS. Sat Nov 05 23:50:33 GMT 2011 Olly Betts <olly@survex.com> * api/documentvaluelist.h,api/omdatabase.cc,api/valueiterator.cc, backends/Makefile.mk,backends/brass/,backends/chert/, backends/inmemory/inmemory_database.cc, backends/multi/multi_valuelist.cc,backends/multivaluelist.h, backends/valuelist.h,backends/valuestats.h,bin/xapian-check-brass.cc, bin/xapian-check-chert.cc,common/,matcher/valuerangepostlist.h, matcher/valuestreamdocument.h: Move multivaluelist.h, valuelist.h and valuestats.h from common to backends. Sat Nov 05 23:40:49 GMT 2011 Olly Betts <olly@survex.com> * api/,backends/brass/brass_postlist.h,backends/chert/chert_postlist.h, backends/contiguousalldocspostlist.h,backends/database.cc, backends/inmemory/inmemory_database.h, backends/multi/multi_postlist.h,backends/remote/net_postlist.h, backends/remote/remote-database.cc,common/,matcher/: Move emptypostlist.h, leafpostlist.h and postlist.h from common to api. Sat Nov 05 23:31:14 GMT 2011 Olly Betts <olly@survex.com> * backends/Makefile.mk,backends/brass/brass_database.cc, backends/chert/chert_database.cc, backends/contiguousalldocspostlist.h,common/Makefile.mk, common/contiguousalldocspostlist.h: Move contiguousalldocspostlist.h from common to backends. Sat Nov 05 23:28:49 GMT 2011 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/Makefile.mk,backends/alltermslist.h, backends/brass/,backends/chert/, backends/inmemory/inmemory_alltermslist.h,common/Makefile.mk, common/alltermslist.h,common/multialltermslist.h: Move alltermslist.h from common to backends. Sat Nov 05 23:20:05 GMT 2011 Olly Betts <olly@survex.com> * common/safesysselect.h: Fix typo (FDSET -> FD_SET). Sat Nov 05 23:12:09 GMT 2011 Olly Betts <olly@survex.com> * common/safesysselect.h: Enhance to provide portability for __WIN32__ too. * net/remoteconnection.cc,net/tcpclient.cc: Update to make use of this. Sat Nov 05 22:39:51 GMT 2011 Olly Betts <olly@survex.com> * common/safesysselect.h: Add wrapper around FD_SET() under __WIN32__ which casts the fd parameter to unsigned to avoid a warning on mingw. * net/tcpclient.cc: Remove __WIN32__-specific case here which was working around this warning. Sat Nov 05 21:58:12 GMT 2011 Olly Betts <olly@survex.com> * bin/.gitignore,bin/Makefile.mk,bin/xapian-delve.cc, docs/admin_notes.rst,examples/,xapian-core.spec.in: Move delve from examples to bin and rename to xapian-delve. Sat Nov 05 15:18:37 GMT 2011 Olly Betts <olly@survex.com> * docs/overview.rst: Fix a few things which were out of date and improve wording in various places. Sat Nov 05 12:07:14 GMT 2011 Olly Betts <olly@survex.com> * docs/quickstart.rst: Fix some factual errors, and tweak .rst source layout a little (":\n::\n" -> "::\n"). Thu Nov 03 20:53:49 GMT 2011 Richard Boulton <richard@tartarus.org> * common/msvc_posix_wrapper.cc,common/safe.cc,net/tcpclient.cc: Some fixes for warnings when cross-compiling with mingw; missing includes to get declarations of functions in windows support code, and change to use the windows SOCKET type to avoid a signed/unsigned comparison warning in tcpclient. Sun Oct 30 23:31:09 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Sun Oct 30 13:16:12 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: List which flags FLAG_DEFAULT includes in the doxygen documentation comments. Sat Oct 29 14:47:04 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Add note to API docs to discourage use of percentage scores. Sat Oct 29 14:40:39 GMT 2011 Olly Betts <olly@survex.com> * docs/quickstart.rst,docs/quickstartexpand.cc.html, docs/quickstartsearch.cc.html,examples/quest.cc, examples/simpleexpand.cc,examples/simplesearch.cc: Use get_weight() instead of get_percent() in examples, since percentages are rather old fashioned and you probably don't want to use them in new applications. Sat Oct 29 13:29:37 GMT 2011 Olly Betts <olly@survex.com> * api/omenquire.cc,common/omenquireinternal.h,docs/deprecation.rst, include/xapian/enquire.h,include/xapian/types.h,tests/api_anydb.cc: Deprecate Xapian::percent and use int instead in the API and our own code. Sat Oct 29 12:30:05 GMT 2011 Olly Betts <olly@survex.com> * docs/quickstart.rst: Fix indentation of example code. Sat Oct 29 10:04:24 GMT 2011 Olly Betts <olly@survex.com> * docs/overview.rst: Fix typo in example code. Fri Oct 28 14:39:09 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/database.h: Note that commit() was new in 1.1.0. Fri Oct 14 12:38:45 GMT 2011 Olly Betts <olly@survex.com> * common/realtime.h: Use safesysselect.h. Mon Oct 03 00:53:13 GMT 2011 Olly Betts <olly@survex.com> * queryparser/lemon.c: Incorporate changes from latest upstream version. I'm failing to follow branch structure upstream to work out which changesets to merge so it's simpler to just compare the latest version with what we currently have. Sun Oct 02 23:31:52 GMT 2011 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge https://www.sqlite.org/src/info/d8bab8cf0b: 'Added %expect directive, to consider a certain number of conflicts "correct." This has the side effect of changing the process exit code to never overflow.' Sun Oct 02 23:26:34 GMT 2011 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge https://www.sqlite.org/src/info/e6cbe1e5ee: "Added -T option, to specify a template filename on the command line. The default is still "lempar.c", though." Not currently useful to us, but trying to minimise differences with upstream lemon. Sun Oct 02 23:14:32 GMT 2011 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge https://www.sqlite.org/src/info/e22c090f35: "Another attempt at fixing the table generator in lemon. Again, this does not effect the SQLite grammar." No change to generated grammar for QueryParser. Sun Oct 02 22:59:28 GMT 2011 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge https://www.sqlite.org/src/info/077a6bee2d: "Fix an issue with lemon generating incorrect grammars. This issue does not effect SQLite." This means we now get the same tables generated as before the previous change. Sun Oct 02 22:25:54 GMT 2011 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Merge https://www.sqlite.org/src/info/27d8e684db: Enhancements to lemon to generate more compact action tables and to avoid making array bounds tests that can never fail on action table calculations. This reduces the size of QueryParser's tables a little. Sun Oct 02 21:47:36 GMT 2011 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge https://www.sqlite.org/src/info/d66a0f31eb: Adjust the lemon implementation so that it always computes the same PDA regardless of qsort() implementation on the host platform. In other words, make all sorts in lemon stable. Thu Sep 29 12:11:28 GMT 2011 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add cases to ensure a quoted boolean prefix accepts an empty value or a single space. Wed Sep 14 04:48:57 GMT 2011 Olly Betts <olly@survex.com> * docs/quickstart.rst: Correct link which was to quickstartsearch.cc.html but should be to quickstartindex.cc.html. Tue Sep 13 08:51:34 GMT 2011 Olly Betts <olly@survex.com> * backends/flint_lock.cc: Fix comment typo. Thu Sep 01 12:50:43 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Clarify "sort by date" with BoolWeight example. Tue Aug 30 13:51:48 GMT 2011 Olly Betts <olly@survex.com> * docs/index.rst: Add link to apidoc.pdf. Wed Aug 24 14:12:05 GMT 2011 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix memory leak (caught by existing testcase queryparser1 when run under valgrind). Wed Aug 24 12:47:49 GMT 2011 Olly Betts <olly@survex.com> * queryparser/,tests/queryparsertest.cc,tests/termgentest.cc: Add support for indexing and searching CJK text using n-grams. Currently this is only enabled if environmental variable XAPIAN_CJK_NGRAM is set to a non-empty value. Tue Aug 23 03:49:27 GMT 2011 Olly Betts <olly@survex.com> * api/vectortermlist.cc,api/vectortermlist.h: Change so that p == NULL represents at_end, and p == data.data() means we've yet to start. Mon Aug 22 15:30:51 GMT 2011 Olly Betts <olly@survex.com> * api/vectortermlist.cc: Add new source file missed from last commit. Mon Aug 22 14:32:50 GMT 2011 Olly Betts <olly@survex.com> * api/,backends/brass/brass_synonym.cc,backends/chert/chert_synonym.cc, common/Makefile.mk,common/vectortermlist.h, queryparser/queryparser.cc: Reimplement VectorTermList to store the list of terms encoded into a single std::string, which is a lot more memory efficient that std::vector<std::string>. Mon Aug 22 12:22:49 GMT 2011 Olly Betts <olly@survex.com> * common/Makefile.mk,common/const_database_wrapper.cc, common/const_database_wrapper.h,matcher/Makefile.mk, matcher/const_database_wrapper.cc,matcher/const_database_wrapper.h: Move const_database_wrapper.h and const_database_wrapper.cc from common/ to matcher/. Sun Aug 21 12:43:38 GMT 2011 Olly Betts <olly@survex.com> * Makefile.am: Since we require GNU find for the coverage-check target, make use of its -delete option, so we no longer also need GNU xargs. Thu Aug 18 06:28:11 GMT 2011 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Reorder header includes to be more standard. Thu Aug 18 01:06:00 GMT 2011 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: <cstring> not <string.h>. Tue Aug 16 04:56:20 GMT 2011 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Throw Xapian::FeatureUnavailableError if backends are disabled, rather than const char *. Tue Aug 16 04:49:05 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_check.cc,backends/brass/brass_check.h, backends/chert/chert_check.cc,backends/chert/chert_check.h, bin/xapian-check.cc: Instead of giving cryptic numeric codes for B-tree errors and then throwing const char *, throw Xapian::DatabaseError() exceptions with string descriptions. Fri Aug 12 23:11:51 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Update from 1.2.7 and ChangeLog. Fri Aug 12 06:38:45 GMT 2011 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc: Factor out throw_bad_message(). Fri Aug 12 06:22:14 GMT 2011 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Factor out throw_database_closed(). Fri Aug 12 06:17:43 GMT 2011 Olly Betts <olly@survex.com> * net/remoteserver.cc: Factor out throw_read_only() function. Fri Aug 12 06:17:01 GMT 2011 Olly Betts <olly@survex.com> * common/noreturn.h: Add comment showing how to use XAPIAN_NORETURN macro. Sat Aug 06 05:15:53 GMT 2011 Olly Betts <olly@survex.com> * api/positioniterator.cc,api/postingsource.cc,api/valueiterator.cc, common/serialise.h,include/xapian/positioniterator.h, include/xapian/postingsource.h: Fix new warnings from -Wshadow with GCC 4.6. These warnings are when a variable name "shadows" a typename, which doesn't seem problematic in general, but the only alternative seems to be to disable -Wshadow entirely, and it is useful in other cases. Thu Aug 04 07:57:09 GMT 2011 Olly Betts <olly@survex.com> * backends/dbfactory_remote.cc,docs/deprecation.rst, include/xapian/dbfactory.h,include/xapian/types.h: Deprecate Xapian::timeout in favour of POSIX type useconds_t. Thu Aug 04 04:04:45 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/termgenerator.h, queryparser/termgenerator_internal.cc: Rename weight parameter to wdf_inc, which much better describes what it does, and avoids a -Wshadow warning with GCC 4.6. Thu Aug 04 04:03:37 GMT 2011 Olly Betts <olly@survex.com> * api/registry.cc: Rename weight variable to weighting_scheme to avoid -Wshadow warning. Wed Aug 03 23:13:36 GMT 2011 Olly Betts <olly@survex.com> * configure.ac: -Wshadow gives bogus warnings with 4.0 (at least on Mac OS X), so disable it for GCC < 4.1 (like the comments suggest we already did!) Wed Aug 03 22:53:55 GMT 2011 Olly Betts <olly@survex.com> * configure.ac: Enable -fshow-column for GCC - things like vim's quickfix mode will then jump to the appropriate column for a compiler error or warning, not just the appropriate line. Wed Aug 03 22:13:26 GMT 2011 Olly Betts <olly@survex.com> * api/omdocument.cc,backends/brass/brass_database.cc, backends/chert/chert_database.cc,common/document.h: Document objects now track if term positions might have been modified, so we can avoid having to consider them at all if you take a document and add or remove a boolean filter term. This gives an 18% speedup for adding tags in notmuch. Mon Aug 01 15:15:17 GMT 2011 Olly Betts <olly@survex.com> * api/Makefile.mk,api/ompostlistiterator.cc,api/postingiterator.cc, include/xapian/postingiterator.h,tests/api_db.cc: Redo PostingIterator based on TermIterator with efficient end iterator checks. Mon Aug 01 13:34:44 GMT 2011 Olly Betts <olly@survex.com> * api/,backends/brass/brass_database.cc, backends/chert/chert_database.cc, backends/inmemory/inmemory_database.cc,common/positionlist.h, examples/delve.cc,include/xapian/positioniterator.h, include/xapian/termiterator.h,include/xapian/valueiterator.h: Take the old SVN end-iterator-proxies branch and rework the TermIterator and PositionIterator reimplementations there to use the same approach as ValueIterator now uses. Sun Jul 31 15:24:55 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/valueiterator.h: Fix comment typo. Sun Jul 31 14:45:46 GMT 2011 Olly Betts <olly@survex.com> * api/omenquire.cc,tests/api_anydb.cc: Previously, Enquire::get_matching_terms_begin() threw InvalidArgumentError if the query was empty. Now we just return an end iterator, which is more consistent with how empty queries behave elsewhere. Sun Jul 31 13:44:06 GMT 2011 Olly Betts <olly@survex.com> * api/valueiterator.cc,include/xapian/valueiterator.h: Rename private ValueIterator::deref() method to decref() (since dereference has another meaning to the intended "remove a reference"). Sun Jul 31 11:27:33 GMT 2011 Olly Betts <olly@survex.com> * api/valueiterator.cc,include/xapian/database.h, include/xapian/document.h,include/xapian/valueiterator.h: Don't use intrusive_ptr to reference count ValueIterator internals, but instead update the reference count in hand-written code. This is less good in itself, but it means we can inline ValueIterator's default constructor and its destructor in the case where the internal is NULL, which should mean a comparison with the end iterator optimises to a NULL pointer check but without needing the ValueIteratorEnd_ class, which doesn't work as ValueIterator() would in templates or some cases of overload resolution. Sat Jul 30 12:07:52 GMT 2011 Olly Betts <olly@survex.com> * docs/doxygen_api.conf.in: Don't generate XML from doxygen for the bindings - the bindings now do this for themselves. (ticket#262) Fri Jul 29 01:33:37 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Just forward declare "class KeyMaker" rather than including <xapian/keymaker.h>. Fri Jul 29 01:28:08 GMT 2011 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/slowvaluelist.cc, include/xapian/database.h,matcher/valuestreamdocument.cc: Change Database::get_document_lazily_() to return void* so we don't need to include xapian/document.h from xapian/database.h just to get Document::Internal declared. This doesn't affect user code, but reduces the number of files which need to be rebuilt in the library when xapian/document.h changes. Wed Jul 27 05:59:32 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/matchspy.h: Remove unnecessary header includes. Wed Jul 27 02:11:08 GMT 2011 Olly Betts <olly@survex.com> * docs/index.rst: Add links to Omega and bindings docs. Tue Jul 26 05:36:50 GMT 2011 Olly Betts <olly@survex.com> * docs/deprecation.rst: Add Omega's OLDP CGI parameter. Fri Jul 22 15:37:48 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Reword documentation of add_boolean_prefix()'s exclusive parameter to talk about terms and prefixes rather than values and fields (which was confusing since "document value" has a particular meaning in Xapian). Thu Jul 21 03:37:49 GMT 2011 Olly Betts <olly@survex.com> * matcher/mergepostlist.cc: Another debug logging fix. Thu Jul 21 03:16:39 GMT 2011 Olly Betts <olly@survex.com> * api/omenquire.cc,api/ompostlistiterator.cc, backends/brass/brass_database.cc, backends/brass/brass_positionlist.cc, backends/brass/brass_postlist.cc,backends/chert/chert_database.cc, backends/chert/chert_positionlist.cc, backends/chert/chert_postlist.cc: More debug logging fixes. Thu Jul 21 02:55:00 GMT 2011 Olly Betts <olly@survex.com> * api/omdatabase.cc,api/omenquire.cc,api/replication.cc, backends/brass/brass_alldocspostlist.cc, backends/brass/brass_database.cc, backends/brass/brass_spellingwordslist.cc,backends/chert/, matcher/mergepostlist.cc,matcher/msetpostlist.cc, matcher/synonympostlist.cc: Fix various incorrect LOGCALL lines. Thu Jul 21 02:12:25 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_postlist.cc,backends/chert/chert_postlist.cc, common/documentterm.h,net/progclient.cc: Fix uses of LOGCALL_VOID which should be LOGCALL_CTOR. Wed Jul 20 07:00:48 GMT 2011 Olly Betts <olly@survex.com> * tests/api_matchspy.cc: Remove checks for remote dbtype in matchspy2 and matchspy4 testcases - these were checking for values which would never match, and the testcases should and do work with remote backends. Wed Jul 20 01:19:46 GMT 2011 Olly Betts <olly@survex.com> * docs/facets.rst: Minor clarification. Wed Jul 20 00:22:38 GMT 2011 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Document in --help that checking a whole database performs additional cross-checks between the tables. Tue Jul 19 18:29:51 GMT 2011 Richard Boulton <richard@tartarus.org> * docs/facets.rst: Expand descriptions for indexing and finding facets to hopefully be more helpful to newcomers. Tue Jul 19 17:52:25 GMT 2011 Richard Boulton <richard@tartarus.org> * docs/facets.rst: Fix example calls to add_matchspy() to be passed pointers, and example call to get_termfreq(). Sat Jul 16 10:54:39 GMT 2011 Olly Betts <olly@survex.com> * tests/collate-test: Now understand use of "#if 0" to disable testcases. * tests/api_db.cc: Remove comment in front of "DEFINE_TESTCASE(multierrhandler1, [...]" which is no longer needed. Wed Jul 13 15:07:27 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/database.h: Improve the documentation comment for Database::close(). (ticket#504) Sat Jul 09 14:00:08 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/stem.h,languages/stem.cc: Stem::operator= now returns a reference to the assigned-to object. Sat Jul 09 07:08:03 GMT 2011 Olly Betts <olly@survex.com> * languages/compiler/analyser.c: Change reporting of line number for repeated string in among to use "FILE:LINE:" prefix. Sat Jul 09 01:04:18 GMT 2011 Olly Betts <olly@survex.com> * common/pretty.h: Don't dereference a NULL pointer. Fri Jul 08 14:28:42 GMT 2011 Olly Betts <olly@survex.com> * common/output.h,common/pretty.h: Fix invalid templates in pretty.h. Fri Jul 08 12:55:19 GMT 2011 Olly Betts <olly@survex.com> * common/pretty.h: Fix template for pretty printing a std::list. Thu Jul 07 10:40:53 GMT 2011 Olly Betts <olly@survex.com> * api/omenquire.cc: Add assertions that the index is in range when dereferencing MSetIterator and ESetIterator. Wed Jul 06 08:20:56 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/chert/chert_table.cc: Trust new to throw std::bad_alloc rather than returning NULL. Wed Jul 06 05:50:38 GMT 2011 Olly Betts <olly@survex.com> * tests/internaltest.cc: Remove testcase serialisequery1. * tests/api_serialise.cc: Extend serialise_query1 to cover equivalent cases via the API. Tue Jul 05 01:20:59 GMT 2011 Olly Betts <olly@survex.com> * common/debuglog.h: Add QUERY category for debug logging. Fri Jul 01 04:55:02 GMT 2011 Olly Betts <olly@survex.com> * common/unordered_map.h: Add my <unordered_map> portability wrapper which was used by the JNI java bindings. It's not currently used here, but is likely to be useful (two of the GSoC projects are using unordered_map). Wed Jun 29 12:16:59 GMT 2011 Richard Boulton <richard@tartarus.org> * languages/compiler/tokeniser.c: Fix memory leak (of file name) in snowball compiler introduced by last commit. Wed Jun 29 07:49:36 GMT 2011 Olly Betts <olly@survex.com> * languages/compiler/: Report "FILE:LINE:" before each error so tools like vim's quickfix mode can parse this and bring up the line with the error automatically. Wed Jun 29 07:49:10 GMT 2011 Olly Betts <olly@survex.com> * languages/compiler/header.h: Add comments to note that struct input must be a prefix of struct tokeniser. Wed Jun 29 07:47:30 GMT 2011 Olly Betts <olly@survex.com> * languages/basque.sbl: Use stringdef instead of literal accented characters in the code. Wed Jun 29 07:13:30 GMT 2011 Olly Betts <olly@survex.com> * docs/stemming.rst,include/xapian/stem.h,languages/: Add stemmers for Armenian (hy), Basque (eu), and Catalan (ca). Mon Jun 27 08:16:55 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/query.h: Add a fake specialised form of the templated ctor for SWIG. Sat Jun 25 11:27:00 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Fix doc comment typo which reversed the intended sense. Sat Jun 25 06:20:34 GMT 2011 Olly Betts <olly@survex.com> * api/omenquire.cc,common/omenquireinternal.h: Remove unused method Enquire::Internal::register_match_decider(). Sat Jun 25 06:05:22 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/enquire.h: MSet(MSet::Internal *) -> MSet(Internal *) for consistency with other classes. Fri Jun 24 16:15:00 GMT 2011 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/slowvaluelist.cc, include/xapian/database.h,matcher/valuestreamdocument.cc: Rename Database::get_document_lazily() to get_document_lazily_() to conform to our convention for naming methods for internal use which have "public" visibility. Wed Jun 22 23:51:15 GMT 2011 Richard Boulton <richard@tartarus.org> * configure.ac: Correct tiny typo in error message when zlib isn't found. Tue Jun 21 04:08:29 GMT 2011 Olly Betts <olly@survex.com> * common/autoptr.h: Update comment about why we have this file. It's no longer a concern that we might need our own AutoPtr implementation but we are likely to want to be able to migrate to unique_ptr easily as C++0x features become more widely available. Tue Jun 21 02:26:36 GMT 2011 Olly Betts <olly@survex.com> * HACKING,common/pretty.h,configure.ac,tests/internaltest.cc: Fix a few lingering references to RefCntPtr and RefCntBase. Tue Jun 21 02:01:25 GMT 2011 Olly Betts <olly@survex.com> * api/,backends/brass/,backends/chert/,backends/database.cc, backends/inmemory/inmemory_alltermslist.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/multi/multi_alltermslist.cc, backends/multi/multi_valuelist.cc,backends/remote/,common/, docs/doxygen_api.conf.in,include/Makefile.mk,include/xapian/, matcher/msetpostlist.h,matcher/multimatch.cc, queryparser/queryparser_internal.h, queryparser/termgenerator_internal.h,tests/internaltest.cc: Convert to use xapian/intrusive_ptr.h instead of xapian/base.h. Mon Jun 20 23:55:10 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/intrusive_ptr.h: Unmodified version of boost/smart_ptr/intrusive_ptr.hpp as of Boost 1.42 (committed to aid merging changes in the future). Mon Jun 20 12:46:47 GMT 2011 Olly Betts <olly@survex.com> * common/databasereplicator.h,languages/steminternal.h,tests/harness/: Remove inclusions of xapian/base.h from files which don't use RefCntPtr or RefCntBase. Mon Jun 20 11:33:03 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/base.h: Remove unnecessary check for self-assignment as it isn't a common case and the current code copes gracefully with it anyway. Mon Jun 20 05:25:05 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_postlist.cc,backends/brass/brass_postlist.h, backends/chert/chert_postlist.cc,backends/chert/chert_postlist.h: Reduce size of BrassPostList/ChertPostList objects by moving bool members together (168->160 bytes for ChertPostList on x86-64 Linux). Sun Jun 19 13:44:32 GMT 2011 Olly Betts <olly@survex.com> * tests/internaltest.cc: Fix typo in comment and verbose output. Sun Jun 19 12:23:43 GMT 2011 Olly Betts <olly@survex.com> * api/error.cc,generate-exceptions: Rearrange members of Xapian::Error to reduce its size (from 48 to 40 bytes on x86-64 Linux). Sun Jun 19 12:22:53 GMT 2011 Olly Betts <olly@survex.com> * common/pretty.h: Fix comment typo. Sun Jun 19 04:45:26 GMT 2011 Olly Betts <olly@survex.com> * INSTALL: Update GCC details - we now recommend 4.3 or newer (was 4.1) and note that while 3.1 is the hard minimum requirement, the oldest we've tested with at all recently was 3.3. Sat Jun 18 14:27:14 GMT 2011 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony, queryparser/termgenerator_internal.cc: Fix warning from GCC 3.3. Sat Jun 18 06:27:03 GMT 2011 Olly Betts <olly@survex.com> * api/,expand/ortermlist.cc,include/xapian/: Use TermIterator() instead of TermIterator(NULL), and similarly for other Xapian iterator classes. Sat Jun 18 05:16:23 GMT 2011 Olly Betts <olly@survex.com> * tests/zlib-vg.so: Remove file committed accidentally. Sat Jun 18 05:07:56 GMT 2011 Olly Betts <olly@survex.com> * api/,include/xapian/positioniterator.h, include/xapian/termiterator.h: Reimplementations of PositionIterator and TermIterator from the end-iterator-proxies branch. Assignment operators for these classes now return *this rather than void. Sat Jun 18 01:49:02 GMT 2011 Olly Betts <olly@survex.com> * tests/perftest/Makefile.mk: Make sure that perftest isn't run with libeatmydata preloaded. Fri Jun 17 07:39:35 GMT 2011 Olly Betts <olly@survex.com> * docs/deprecation.rst: Tabs to spaces. Fri Jun 17 07:38:06 GMT 2011 Olly Betts <olly@survex.com> * docs/deprecation.rst: Add flush() -> commit(). Thu Jun 16 14:40:35 GMT 2011 Olly Betts <olly@survex.com> * tests/: Add TEST( ) or TEST(! ) around existing reopen() calls to check that we get true or false returned as we would expect. (ticket#548) Thu Jun 16 05:15:28 GMT 2011 Olly Betts <olly@survex.com> * README: Note licence is GPL *2+*. Say "SVN/git". Use trac URL for wiki. Wed Jun 15 23:31:03 GMT 2011 Olly Betts <olly@survex.com> * xapian-core/api/error.cc: Don't include the magic code byte at the start of Error::get_description()'s return value. Wed Jun 15 15:51:42 GMT 2011 Olly Betts <olly@survex.com> * exception_data.pm,generate-exceptions,net/serialise.cc: When propagating exceptions from a remote backend server, use a numeric code to represent which exception is being transferred rather than the name of the type as that can be turned back into an exception with a simple switch statement. It's also less data to transfer. (ticket#471) Wed Jun 15 14:03:49 GMT 2011 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remote-database.h, common/remoteprotocol.h,docs/remote_protocol.rst,net/remoteserver.cc: Propagate the return value of Database::reopen() across the link. (ticket#548) Wed Jun 15 13:03:56 GMT 2011 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remote-database.h, common/remoteprotocol.h,docs/remote_protocol.rst,net/remoteserver.cc: Unify REPLY_GREETING and REPLY_UPDATE; send (last_docid - doccount) instead of last_docid; send (doclen_ubound - doclen_lbound) instead of doclen_ubound. This requires a remote protocol major version bump. Fix out of date documentation for what's in a REPLY_UPDATE message. Wed Jun 15 12:24:07 GMT 2011 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc: Remove special check which gives a more helpful error message when a modern client is used against a remote server running Xapian <= 0.9.6. Wed Jun 15 11:07:19 GMT 2011 Olly Betts <olly@survex.com> * net/remoteserver.cc: Remove unnecessary call to reopen() - either we just called it or we're a writable database and it doesn't do anything. Wed Jun 15 09:03:06 GMT 2011 Olly Betts <olly@survex.com> * xapian-core/backends/remote/remote-database.cc, xapian-core/common/remote-database.h: Factor out the code to decode the returned stats into a new apply_stats_update() method. Wed Jun 15 08:57:12 GMT 2011 Olly Betts <olly@survex.com> * xapian-core/Makefile.am: Remove check-* from .PHONY as that doesn't work with a pattern rule. Fix check-% pattern rule to actually work. Tue Jun 14 15:11:38 GMT 2011 Olly Betts <olly@survex.com> * docs/remote_protocol.rst: Fixed typo which reversed the intended sense. Tue Jun 14 02:01:42 GMT 2011 Olly Betts <olly@survex.com> * Makefile.am,configure.ac: Avoid portability warning from automake about GNU make %-style pattern rules. Tue Jun 14 01:42:01 GMT 2011 Olly Betts <olly@survex.com> * configure.ac: Add unnecessary AC_LANG_SOURCE wrapper to shut up autoconf warning. Tue Jun 14 01:18:54 GMT 2011 Olly Betts <olly@survex.com> * configure.ac: Reset LIBRARY_VERSION_INFO to 0:0:0 for the development series, since the library name is different (it has a -1.3 suffix). Mon Jun 13 16:24:20 GMT 2011 Olly Betts <olly@survex.com> * api/omenquire.cc,common/multimatch.h,common/omenquireinternal.h, docs/deprecation.rst,include/xapian/enquire.h,matcher/multimatch.cc, net/remoteserver.cc,tests/api_backend.cc: Remove the deprecated old-style match spy approach of using a MatchDecider. Mon Jun 13 15:47:57 GMT 2011 Olly Betts <olly@survex.com> * api/keymaker.cc,docs/deprecation.rst,include/xapian/keymaker.h, tests/Makefile.am,tests/api_sortingold.cc: Remove deprecated Sorter class and MultiValueSorter subclass. * tests/Makefile.mk,tests/perftest/Makefile.mk, tests/soaktest/Makefile.mk: Make the generated "*_all.h" depend on the Makefile.am or Makefile.mk which contains the list of files it is generated from so that it gets regenerated when a file is removed from that list. Mon Jun 13 13:18:29 GMT 2011 Olly Betts <olly@survex.com> * common/omenquireinternal.h: Remove a FIXME - caching the Document object in MSetItem isn't a replacement for storing the sort_key now that sort_keys can be computed. Mon Jun 13 13:07:31 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.cc: Fold the two forms of QueryParser::add_boolean_prefix() into one method with a default parameter now that changing the ABI isn't a worry. Mon Jun 13 12:00:36 GMT 2011 Olly Betts <olly@survex.com> * docs/deprecation.rst: Non-pythonic iterators have been removed. (ticket#255) Mon Jun 13 10:40:30 GMT 2011 Olly Betts <olly@survex.com> * docs/deprecation.rst: Python Stem_get_available_languages() has now been removed. Mon Jun 13 10:38:49 GMT 2011 Olly Betts <olly@survex.com> * docs/deprecation.rst,include/xapian/enquire.h: Remove deprecated default value for second parameter to Enquire::set_sort_by_value() and friends. Mon Jun 13 09:30:15 GMT 2011 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/brass/brass_database.cc, backends/brass/brass_database.h,backends/chert/chert_database.cc, backends/chert/chert_database.h,backends/database.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/remote/remote-database.cc,common/, include/xapian/database.h: Database::reopen() now returns true if the database may have been reopened. (ticket#548) Mon Jun 13 09:01:33 GMT 2011 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Add note about xapian-chert-update. Mon Jun 13 08:47:06 GMT 2011 Olly Betts <olly@survex.com> * bin/xapian-check-flint.cc,bin/xapian-check-flint.h: Remove two flint-specific files I missed. Mon Jun 13 08:42:11 GMT 2011 Olly Betts <olly@survex.com> * HACKING,INSTALL,Makefile.am,api/compactor.cc,api/replication.cc, backends/Makefile.mk,backends/brass/brass_compact.cc, backends/chert/chert_compact.cc,backends/databasereplicator.cc, backends/dbfactory.cc,backends/dir_contents,backends/flint/, backends/slowvaluelist.h,bin/,common/database.h,common/pretty.h, common/unaligned.h,configure.ac,docs/,include/xapian/dbfactory.h, include/xapian/valueiterator.h,include/xapian/version_h.cc,tests/, tests/harness/,tests/queryparsertest.cc,tests/termgentest.cc, tests/testdata/flint-0.9.9/,tests/testdata/flint-1.0.1/, tests/testdata/flint-1.0.2/: Remove flint backend. Mon Jun 13 06:09:29 GMT 2011 Olly Betts <olly@survex.com> * bin/Makefile.mk: Remove lingering reference to quartz (rules to generate man pages for quartzdump, etc). Mon Jun 13 05:46:00 GMT 2011 Olly Betts <olly@survex.com> * Makefile.am,tests/Makefile.am: Just use a GNU-make-specific pattern rules to forward check-* targets from the top level to the tests subdirectory. It's clear we aren't keeping the explicit list of target forwarding rules up to date, and this is just a convenience so not worth a lot of maintenance effort. Mon Jun 13 05:01:33 GMT 2011 Olly Betts <olly@survex.com> * configure.ac: Update version to 1.3.0 and add -1.3 suffix for library and /xapian-1.3 suffix to include file installation directory. Mon Jun 13 04:56:13 GMT 2011 Olly Betts <olly@survex.com> * unicode/tclUniData.cc: Upgrade to Unicode 6.0.0 (ticket#497). * tests/api_unicode.cc: Extend testcases to check for assorted changes and additions in Unicode 6.0.0. Sun Jun 12 11:54:45 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Final update for 1.2.6. Sat Jun 11 10:11:23 GMT 2011 Olly Betts <olly@survex.com> * api/omdocument.cc: Update LOGCALL(MATCH, ...) to LOGCALL(DB, ...). Convert LOGLINE() at start of method to LOGCALL(). Sat Jun 11 09:57:00 GMT 2011 Olly Betts <olly@survex.com> * api/omdatabase.cc: Remove self-assignment check from Database::operator= since this is an uncommon code path and RefCntPtr ensures that self-assignment of the internals is safe. Sat Jun 11 08:48:10 GMT 2011 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/brass/brass_cursor.h, backends/chert/chert_cursor.h,backends/flint/flint_cursor.h, backends/flint/flint_table.cc,backends/flint/flint_values.cc, backends/inmemory/inmemory_database.cc,bin/xapian-check-brass.cc, bin/xapian-check-chert.cc,bin/xapian-chert-update.cc, matcher/queryoptimiser.cc,matcher/valuegepostlist.h: Use string() or std::string() rather than "". * common/remoteconnection.h,net/tcpserver.cc: Use 2 parameter form of Error subclass ctors when the context is "". * api/replication.cc,backends/flint/flint_database.cc, common/remoteconnection.h,net/replicatetcpclient.cc, net/replicatetcpserver.cc: Make context parameter of RemoteConnection ctor optional. * common/leafpostlist.h: Say 'empty' rather than '""' in a documentation comment. * examples/simpleindex.cc: Use '.resize(0)' rather than '= ""'. Sat Jun 11 07:01:41 GMT 2011 Olly Betts <olly@survex.com> * api/,backends/brass/,backends/chert/,backends/flint/, backends/remote/remote-database.cc,common/,include/xapian/,matcher/, net/remoteserver.cc,net/serialise.cc: Consistently use "slot" in variable names for value slots, rather than valueno, valno, valueid or value. The only exception remaining is a protected member of Xapian::StringValueRangeProcessor since that's really part of the public API. Fri Jun 10 11:55:45 GMT 2011 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update in preparation for 1.2.6. Fri Jun 10 09:05:38 GMT 2011 Olly Betts <olly@survex.com> * examples/NEWS,examples/TODO: Remove files which are unused since the examples were merged into xapian-core. Fri Jun 10 09:03:33 GMT 2011 Olly Betts <olly@survex.com> * Makefile.am: Ship ChangeLog.0 in the tarball. Fri Jun 10 05:19:24 GMT 2011 Olly Betts <olly@survex.com> * tests/Makefile.am: Actually include soaktest/Makefile.mk. Thu Jun 09 23:26:52 GMT 2011 Olly Betts <olly@survex.com> * docs/doxygen_api.conf.in,docs/doxygen_source.conf.in: Define XAPIAN_DEPRECATED_CLASS to empty for doxygen. Thu Jun 09 22:21:29 GMT 2011 Olly Betts <olly@survex.com> * docs/overview.rst: Document "remote" in stub databases. Thu Jun 09 15:23:12 GMT 2011 Olly Betts <olly@survex.com> * HACKING,docs/deprecation.rst,include/xapian/deprecated.h, include/xapian/keymaker.h: Add XAPIAN_DEPRECATED_CLASS macro for marking a class as deprecated, so we don't have to call XAPIAN_DEPRECATED() with no parameters. Thu Jun 09 14:03:17 GMT 2011 Olly Betts <olly@survex.com> * HACKING: Clarify wording about PATH. Thu Jun 09 13:04:53 GMT 2011 Olly Betts <olly@survex.com> * docs/queryparser.rst,include/xapian/queryparser.h, queryparser/queryparser.cc,queryparser/queryparser.lemony, queryparser/queryparser_internal.h,tests/queryparsertest.cc: Add QueryParser::set_max_wildcard_expansion() method to allow limiting the number of terms a wildcard can expand to. Implementation mostly from a patch by Adam Sjøgren in ticket#350. Tue Jun 07 23:16:33 GMT 2011 Dan Colish <dcolish@gmail.com> * backends/brass/brass_changesetapplier.h, backends/chert/chert_changesetapplier.h, backends/flint/flint_changesetapplier.h: Remove dead code Thu Jun 02 13:21:28 GMT 2011 Olly Betts <olly@survex.com> * AUTHORS: Add recent bug reporters. Thu Jun 02 11:03:57 GMT 2011 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: If default_op is OP_NEAR or OP_PHRASE then disable stemming of the terms, since we don't index positional information for stemmed terms by default. * tests/queryparsertest.cc: Adjust near1 and phrase1 to act as regression tests for this fix. Thu Jun 02 02:58:58 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_spelling.cc,backends/brass/brass_spelling.h, backends/chert/chert_spelling.cc,backends/chert/chert_spelling.h, backends/flint/flint_spelling.cc,backends/flint/flint_spelling.h: Fix bug Nikita Smetanin spotted with adding a word which has the same trigram in an even number of times. * tests/api_spelling.cc: Add regression test spell8. Wed Jun 01 14:11:38 GMT 2011 Olly Betts <olly@survex.com> * backends/flint_lock.cc: Kill the child process which holds the lock with SIGKILL as that can't be ignored, whereas SIGHUP can be in some cases it seems. Wed Jun 01 12:12:14 GMT 2011 Olly Betts <olly@survex.com> * docs/quickstartexpand.cc.html,docs/quickstartindex.cc.html, docs/quickstartsearch.cc.html: Reinstate the old HTML versions of these files. Wed Jun 01 12:09:55 GMT 2011 Olly Betts <olly@survex.com> * docs/queryparser.rst,docs/remote_protocol.rst: Fix literal hard spaces in input to just be spaces. Wed Jun 01 12:06:26 GMT 2011 Olly Betts <olly@survex.com> * docs/quickstart.rst: Fix sub-heading run in to previous paragraph. Wed Jun 01 12:04:52 GMT 2011 Olly Betts <olly@survex.com> * docs/remote.rst: Replace out-dated list of xapian-tcpsrv's command line options with a reference to --help and the man page. Tweak wording in a few places. Wed Jun 01 11:54:37 GMT 2011 Olly Betts <olly@survex.com> * docs/overview.rst: Change omrset to rset. Remove some bogus `` left over from fixing definition list formatting. Wed Jun 01 11:38:09 GMT 2011 Olly Betts <olly@survex.com> * docs/stemming.rst: Comment out bogus claim that there are stopword lists in xapian-data (the lack of stopword lists is already ticket#269). Wed Jun 01 11:29:58 GMT 2011 Olly Betts <olly@survex.com> * docs/matcherdesign.rst,docs/stemming.rst: Convert more arrows to Unicode. * docs/stemming.rst: Fix some missing paragraph breaks. Wed Jun 01 11:18:25 GMT 2011 Olly Betts <olly@survex.com> * docs/intro_ir.rst: Use Unicode right arrow. Wed Jun 01 10:38:27 GMT 2011 Olly Betts <olly@survex.com> * docs/: Revert quickstart*.cc.html to being HTML rather than .rst as the .rst version lose the syntax highlighting. Update svn:ignore. Wed Jun 01 10:27:11 GMT 2011 Olly Betts <olly@survex.com> * docs/intro_ir.rst: Sort out the remaining formatting issues. Wed Jun 01 08:05:05 GMT 2011 Olly Betts <olly@survex.com> * docs/intro_ir.rst: More markup fixes. Wed Jun 01 06:16:56 GMT 2011 Olly Betts <olly@survex.com> * docs/tests.rst: Use generated contents list instead of manual one. Wed Jun 01 06:16:13 GMT 2011 Olly Betts <olly@survex.com> * docs/index.rst,docs/internals.rst: Centre links at top; Sort out <hr> to be full width. Wed Jun 01 03:23:20 GMT 2011 Olly Betts <olly@survex.com> * docs/bm25.rst: Sort out inline equations. Tue May 31 15:21:10 GMT 2011 Olly Betts <olly@survex.com> * docs/spelling.rst: Add link to Omega documentation. Tue May 31 07:52:19 GMT 2011 Olly Betts <olly@survex.com> * docs/overview.rst,docs/quickstart.rst,docs/remote_protocol.rst: More .rst formatting fixes. Tue May 31 05:02:48 GMT 2011 Olly Betts <olly@survex.com> * docs/overview.rst: More .rst fixes. Tue May 31 04:32:52 GMT 2011 Olly Betts <olly@survex.com> * docs/: More .rst fixes. Tue May 31 04:11:27 GMT 2011 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/code_structure.rst: Revert code_structure.html conversion - this file is generated by a script. Tue May 31 03:56:39 GMT 2011 Olly Betts <olly@survex.com> * docs/install.rst,docs/stemming.rst,docs/tests.rst: Fix up some .rst conversion issues. Tue May 31 03:36:49 GMT 2011 Dan Colish <dcolish@gmail.com> * docs/remote_protocol.rst docs/quickstart.rst docs/remote.rst docs/internals.rst docs/scalability.rst docs/quickstartsearch.cc.rst docs/bm25.rst docs/install.rst docs/quickstartindex.cc.rst docs/index.rst docs/matcherdesign.rst docs/quickstartexpand.cc.rst docs/queryparser.rst docs/stemming.rst docs/intro_ir.rst docs/code_structure.rst docs/Makefile.am docs/tests.rst docs/overview.rst: Bulk port all remaining documentation to RST. Sun May 29 07:04:42 GMT 2011 Olly Betts <olly@survex.com> * docs/postingsource.rst: Add PostingSource example. (ticket#503) Sun May 29 05:19:40 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/database.h: Add @exception InvalidArgumentError for Database::get_document() (ticket#542). Fri May 27 05:36:28 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Thu May 26 02:44:07 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: FIXME:1.1.3 -> FIXME:1.3. Wed May 25 14:58:37 GMT 2011 Olly Betts <olly@survex.com> * docs/queryparser.html: Document the precedence order of operators. Fri May 20 04:13:54 GMT 2011 Olly Betts <olly@survex.com> * examples/simpleindex.cc: Add short description to usage message. Tue May 17 21:44:55 GMT 2011 Olly Betts <olly@survex.com> * docs/scalability.html: Bring up-to-date. Wed May 11 01:34:34 GMT 2011 Olly Betts <olly@survex.com> * AUTHORS: Add Scott Zhang for mingw fixes. Wed May 11 01:30:54 GMT 2011 Olly Betts <olly@survex.com> * net/progclient.cc,net/tcpserver.cc: Need <cstdio> for sprintf on __WIN32__ and cygwin. Wed May 11 01:22:07 GMT 2011 Olly Betts <olly@survex.com> * configure.ac: Define __MSVCRT_VERSION__ to 0x0601 under mingw so we get _ftime64(). Wed May 11 01:19:14 GMT 2011 Olly Betts <olly@survex.com> * common/closefrom.cc,common/closefrom.h: We don't need closefrom() under __WIN32__ currently, so disable it there as it doesn't compile under mingw. Mon May 09 20:13:17 GMT 2011 Olly Betts <olly@survex.com> * examples/delve.cc: Report has_positions(). Thu Apr 07 05:38:36 GMT 2011 Dan Colish <dcolish@gmail.com> * matcher/phrasepostlist.cc, matcher/queryoptimiser.cc, matcher/exactphrasepostlist.cc, matcher/exactphrasepostlist.h, matcher/phrasepostlist.h: Remove temporary vector in queryoptimiser and replace with iterators. Wed Apr 06 10:27:41 GMT 2011 Richard Boulton <richard@tartarus.org> * docs/admin_notes.rst: Remove the word "also", which makes sense when read in context of the previous section, but is confusing if the paragraph is read standalone, and isn't necessary. Mon Apr 04 14:41:33 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Final update for 1.2.5. Mon Apr 04 14:06:56 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_version.cc: Remove ? from comment - the brass version change is definitely going to be in 1.2.5. Mon Apr 04 13:57:23 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Bump release date. Sat Mar 26 14:49:41 GMT 2011 Olly Betts <olly@survex.com> * INSTALL: Fix typo in previous commit. Sat Mar 26 14:03:15 GMT 2011 Olly Betts <olly@survex.com> * INSTALL: Note how to build for a non-default arch on a multi-arch platform. Sat Mar 26 13:22:49 GMT 2011 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update for 1.2.5. Sat Mar 26 12:28:38 GMT 2011 Olly Betts <olly@survex.com> * AUTHORS: Update. Sat Mar 26 10:33:19 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_postlist.cc,backends/brass/brass_postlist.h, backends/chert/chert_postlist.cc,backends/chert/chert_postlist.h, backends/inmemory/inmemory_database.h: Address issues highlighted by clang++ warnings. Fri Mar 25 05:36:41 GMT 2011 Olly Betts <olly@survex.com> * docs/tests.html: Slight tweak to improve up-to-date-ness, but more work would be useful. Fri Mar 25 05:30:56 GMT 2011 Olly Betts <olly@survex.com> * docs/matcherdesign.html: Update - quite a bit has changed in the matcher since this was last updated! Thu Mar 24 23:31:57 GMT 2011 Olly Betts <olly@survex.com> * api/replication.cc,common/replication.h: Pass reader_close_time as double everywhere, rather than truncating it to int. Include reader_close_time parameters in debug logging. Document reader_close_time in doxygen comments. Tue Mar 22 15:36:25 GMT 2011 Dan Colish <dcolish@gmail.com> * backends/brass/brass_database.h: Remove left over copyright. Tue Mar 22 04:54:12 GMT 2011 Dan Colish <dcolish@gmail.com> * backends/brass/brass_btreebase.cc,backends/chert/chert_btreebase.cc, backends/flint/flint_btreebase.cc: Correct comments for which revision is packed when writing base files Tue Mar 22 03:18:19 GMT 2011 Dan Colish <dcolish@gmail.com> * backends/brass/brass_database.cc,backends/chert/chert_database.cc, backends/flint/flint_database.cc,include/xapian/database.h: Remove FIXME and comments for DB_OVERWRITE. Tue Mar 22 01:18:38 GMT 2011 Dan Colish <dcolish@gmail.com> * backends/brass/,backends/chert/chert_database.cc, backends/flint/flint_database.cc,tests/api_replicate.cc: Clean up changesets when replicating. Add test coverage for this feature. Allow XAPIAN_MAX_CHANGESETS to be altered without reopening the database. Closes Ticket #278. Mon Mar 21 01:50:41 GMT 2011 Olly Betts <olly@survex.com> * docs/bm25.html: Add a link to the 1976 Robertson/Sparck Jones paper. Wed Mar 16 05:56:58 GMT 2011 Dan Colish <dcolish@gmail.com> * bin/xapian-check-flint.cc: Initialize did and current_wdf to prevent uninitalized usage and compiler warnings. Mon Mar 14 03:20:18 GMT 2011 Olly Betts <olly@survex.com> * common/debuglog.cc,common/debuglog.h: Rename indent member to indent_level so it doesn't clash with indent() method. Mon Mar 14 02:44:24 GMT 2011 Olly Betts <olly@survex.com> * common/debuglog.cc,common/debuglog.h: Use an integer counter for the indent rather than a std::string which we add/remove spaces to/from. This is cleaner, but also there seems to be an issue with having a std::string member in a global static object on OS X. Fri Mar 11 01:12:04 GMT 2011 Olly Betts <olly@survex.com> * HACKING: Document using '{ }' rather than ';' for empty loop bodies. Fri Mar 11 01:02:47 GMT 2011 Olly Betts <olly@survex.com> * HACKING: Remove bogus ; after method definition in code example. Fri Mar 11 00:07:22 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc,backends/chert/chert_database.cc, backends/flint/flint_database.cc: Factor out literal constant as MAX_OPEN_RETRIES. Wed Mar 09 14:54:06 GMT 2011 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Improve the new testcases a bit. Wed Mar 09 14:11:59 GMT 2011 Olly Betts <olly@survex.com> * api/omenquire.cc,common/esetinternal.h,common/omenquireinternal.h, expand/esetinternal.cc,expand/expandweight.cc, include/xapian/enquire.h,tests/api_anydb.cc: Enquire::get_eset() now accepts a min_wt argument to allow the minimum wanted weight to be specified. Default is 0, which gives the previous behaviour. Wed Mar 09 00:00:25 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/termiterator.h: Fix misspelling in doc comment. Tue Mar 08 06:58:14 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/brass/brass_table.h, backends/chert/chert_table.cc,backends/chert/chert_table.h, backends/flint/flint_table.cc,backends/flint/flint_table.h: Use defined constant BLOCK_CAPACITY instead of hardcoded 4 (ticket#536). Tue Mar 08 05:56:42 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/chert/chert_table.cc, backends/flint/flint_table.cc: If we try to delete an old base file and it isn't there, just continue rather than throwing an exception. We wanted to get rid of it anyway, and it may be NFS issues telling us the wrong thing. In particular, DatabaseCoruptError was rather a pessimistic assessment. Tue Mar 08 05:55:48 GMT 2011 Olly Betts <olly@survex.com> * common/io_utils.h: Note that io_unlink() may return false when it should have returned true on NFS. Sun Mar 06 23:25:47 GMT 2011 Olly Betts <olly@survex.com> * common/io_utils.cc,common/io_utils.h: Add io_unlink(). * backends/brass/brass_database.cc,backends/brass/brass_table.cc, backends/chert/chert_database.cc,backends/chert/chert_table.cc, backends/flint/flint_database.cc,backends/flint/flint_table.cc: Use io_unlink() instead of sys_unlink_if_exists(). Sun Mar 06 22:26:36 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc,backends/chert/chert_database.cc, backends/flint/flint_database.cc: If DANGEROUS mode is turned on, then actually set the flag for this in replication changes files (the reader will currently throw an exception, but that's better than quietly handling them incorrectly). Sat Mar 05 12:58:31 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Sat Mar 05 12:49:16 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Fix a few typos. Sat Mar 05 06:19:25 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Fix typo in old entry. Sat Mar 05 06:16:41 GMT 2011 Olly Betts <olly@survex.com> * common/remoteconnection.h: Internal doc comment improvements: Fix typo, finish truncated sentence, wrap lines to 80 columns. Sat Mar 05 02:59:35 GMT 2011 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc: Make sure port number is specified. Add "(required)" after --host and --port in help. Sat Mar 05 02:50:04 GMT 2011 Olly Betts <olly@survex.com> * docs/replication.rst: Mention new defaulting of -m. Sat Mar 05 02:44:37 GMT 2011 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc: If --master isn't specified, default to DATABASE. Sat Mar 05 00:14:21 GMT 2011 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc: Give an error if the host isn't set, rather than trying to connect to an empty hostname, which gives the error "Couldn't resolve host (Unknown server error)", which might confuse if you fail to notice the double space and realise what it means. Fri Mar 04 17:20:13 GMT 2011 Richard Boulton <richard@tartarus.org> * docs/replication.rst: Update documentation to make it clear that users shouldn't try to create the destination directory for replication themselves. Thu Mar 03 11:10:10 GMT 2011 Olly Betts <olly@survex.com> * docs/intro_ir.html: Fix typo in author's name. Thu Mar 03 11:06:30 GMT 2011 Olly Betts <olly@survex.com> * docs/intro_ir.html: Update link to a paper. Update text about book "to be published in 2008". Tue Mar 01 11:39:40 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Mon Feb 28 12:40:24 GMT 2011 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Avoid leaking postlist tree if an exception is thrown during the match. Wed Feb 23 15:26:56 GMT 2011 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/index.html: Process collapsing.rst and link it into the documentation. Wed Feb 23 15:22:06 GMT 2011 Olly Betts <olly@survex.com> * docs/collapsing.rst: Add missing document (for some reason this file was empty in SVN, but present in my tree under a different name, and not checked in). Mon Feb 21 14:49:06 GMT 2011 Olly Betts <olly@survex.com> * docs/deprecation.rst: Deprecate MSet.items and ESet.items. Mon Feb 21 14:06:21 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc,backends/chert/chert_database.cc, backends/flint/flint_database.cc,common/remoteconnection.h, net/remoteconnection.cc: Pass a file descriptor to RemoteConnection::send_file() to avoid a race between the caller checking if a file exists and send_file() trying to open it to send it. Mon Feb 21 13:50:06 GMT 2011 Olly Betts <olly@survex.com> * queryparser/lemon.c: Fix issues detected by DACA cppcheck run on the Debian archive. Mon Feb 21 13:30:55 GMT 2011 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Fix spaceterms1 which was never running one part of the testcase (and had been that way for years). Fix an off-by-one error in the enabled code. Mon Feb 21 13:05:36 GMT 2011 Olly Betts <olly@survex.com> * tests/api_backend.cc: Add some test coverage for DatabaseModifiedError in get_mset(). Mon Feb 21 12:03:35 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Fix typo in old entry. Mon Feb 21 11:36:07 GMT 2011 Olly Betts <olly@survex.com> * examples/quest.cc: Report any spelling correction (requires the database contains spelling data of course). Wed Feb 16 13:11:03 GMT 2011 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Add doxygen markup so alternative overloaded forms of Enquire::get_mset() appear in the API documentation. Wed Feb 09 14:24:57 GMT 2011 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Up to darte for 1.2.5. Minor wording improvements. Mention copydatabase --no-renumber. Wed Feb 09 14:06:55 GMT 2011 Olly Betts <olly@survex.com> * examples/copydatabase.cc: Add --no-renumber option. Tue Feb 01 11:17:58 GMT 2011 Olly Betts <olly@survex.com> * configure.ac: -Wstrict-null-sentinel was added in GCC 4.0.1 so doesn't work with GCC 4.0.0. For simplicity, only enable it for GCC >= 4.1. Tue Feb 1 08:22:52 GMT 2011 Olly Betts <olly@survex.com> * docs/deprecation.rst: Add $set{spelling,true}. Thu Jan 20 14:52:53 GMT 2011 Olly Betts <olly@survex.com> * backends/flint/flint_btreebase.cc,backends/flint/flint_btreebase.h, backends/flint/flint_table.cc: Don't read the bitmaps from the base files when opening a database for reading (cross-port of equivalent change to chert). Thu Jan 20 14:37:10 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_btreebase.cc,backends/brass/brass_btreebase.h, backends/brass/brass_table.cc: Don't read the bitmaps from the base files when opening a database for reading (cross-port of equivalent change to chert). Thu Jan 20 14:21:10 GMT 2011 Olly Betts <olly@survex.com> * backends/chert/chert_btreebase.cc,backends/chert/chert_btreebase.h, backends/chert/chert_table.cc: Don't read the bitmaps from the base files when opening a database for reading. Thu Jan 20 01:57:02 GMT 2011 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc: Optimise not to update doclength when it hasn't changed (cross-port of equivalent change to chert). Thu Jan 20 01:36:51 GMT 2011 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc: Optimise not to update doclength when it hasn't changed. Wed Jan 19 03:50:18 GMT 2011 Olly Betts <olly@survex.com> * tests/api_compact.cc: Close ofstream objects used to write out stub files for testing before we run the actual compaction, to avoid issues on Microsoft Windows (ticket#525). Wed Jan 19 01:01:38 GMT 2011 Olly Betts <olly@survex.com> * HACKING: Snapshots and releases are now bootstrapped with autoconf 2.68 and libtool 2.4. Prune information about reasons for needing really old autotools versions when there's a reason to need a newer version anyway. Sat Jan 15 11:10:26 GMT 2011 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog and 1.0.23. Sat Jan 15 10:28:10 GMT 2011 Olly Betts <olly@survex.com> * bin/xapian-chert-update.cc: Fix to handle value slot entries in the termlist table. Wed Jan 12 23:49:21 GMT 2011 Olly Betts <olly@survex.com> * bin/xapian-chert-update.cc: Adjust keys for doclength chunks too. Wed Jan 12 13:37:30 GMT 2011 Olly Betts <olly@survex.com> * AUTHORS,bin/xapian-chert-update.cc: Fix to also rewrite docids in value chunk keys (reported by Luca Barbieri on xapian-discuss). Tue Jan 11 08:41:02 GMT 2011 Olly Betts <olly@survex.com> * docs/deprecation.rst: Note removal of if idx in mset. Tue Jan 11 08:30:29 GMT 2011 Olly Betts <olly@survex.com> * docs/deprecation.rst: PostingSource now offers a replacement for Enquire::set_bias(). Mon Jan 10 09:40:24 GMT 2011 Olly Betts <olly@survex.com> * api/compactor.cc: Add missing header <ctime> for time() (ticket#530). Sun Jan 09 23:14:02 GMT 2011 Olly Betts <olly@survex.com> * api/compactor.cc: Use msvc_posix_rename() under __WIN32__ to atomically update stub file after compaction (ticket#525). Thu Dec 23 13:00:43 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Whitespace consistency tweak. Thu Dec 23 11:01:26 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Handle NEAR/<offset> and ADJ/<offset> where offset isn't an integer the same way at the end of the query as in the middle. Tue Dec 21 10:28:08 GMT 2010 Olly Betts <olly@survex.com> * net/tcpserver.cc: If we can't bind to the specified port because it is a privileged one, exit with code 77 (EX_NOPERM) to make it easier to automatically handle failure when starting the server from a script. Tue Dec 21 07:43:52 GMT 2010 Olly Betts <olly@survex.com> * docs/index.html: Add link to main website. Tue Dec 21 07:40:57 GMT 2010 Olly Betts <olly@survex.com> * docs/index.html,docs/overview.html: Update links to wiki.xapian.org to point to trac.xapian.org/wiki instead. Mon Dec 20 10:02:06 GMT 2010 Richard Boulton <richard@tartarus.org> * docs/deprecation.rst: Add note about botched removal of python's Enquire.get_matching_terms (now fully removed). Sun Dec 19 12:40:12 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update release date. Sat Dec 18 13:17:52 GMT 2010 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update for 1.2.4. Wed Dec 15 11:56:22 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update. Tue Dec 14 12:46:36 GMT 2010 Olly Betts <olly@survex.com> * HACKING,tests/runtest.in: Add XAPIAN_TESTSUITE_LD_PRELOAD hook to allow libeatmydata to easily be used when running the testsuite. Mon Dec 13 14:28:19 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/: Make it more explicitly clear that PostingIterator, PositionIterator, and TermIterator's skip_to methods advance (and hence shouldn't be expected to allow you to "rewind" the stream). Mon Dec 13 14:18:18 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/database.h: "network databases" -> "remote databases". Mon Dec 13 14:10:08 GMT 2010 Olly Betts <olly@survex.com> * net/remoteserver.cc: If the message parameter only contains a string then we can just use it as is. Mon Dec 13 13:34:30 GMT 2010 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remote-database.h, common/remoteprotocol.h,common/remoteserver.h, docs/remote_protocol.html,net/remoteserver.cc,tests/api_metadata.cc: Add support for iterating metadata keys with the remote backend. This change necessitated a minor version bump in the remote protocol. Sun Dec 12 12:24:48 GMT 2010 Olly Betts <olly@survex.com> * weight/tradweight.cc: Fix calculation order to avoid inconsistent weights due to rounding for TradWeight(0). * tests/api_backend.cc: Add regression test tradweight2. Sat Dec 11 11:38:32 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS: Update thanks list with bug reporters and patch submitters. Thu Dec 09 03:59:42 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Thu Dec 09 03:58:20 GMT 2010 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Add FIXME note about using sendfile() or similar. Tue Dec 07 10:16:31 GMT 2010 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Note 1.0.x version #515 was fixed in. Mon Dec 06 05:26:01 GMT 2010 Olly Betts <olly@survex.com> * net/replicatetcpclient.cc: Rearrange loop to avoid duplicating code. Mon Dec 06 01:39:23 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix typo so we test for OP_NEAR or OP_PHRASE, not twice for OP_NEAR, which fixes a bug with not setting the correct window size for default_op of OP_PHRASE in some cases. Factor out the "is_positional(op)" test into an inlined function to help avoid repeating this error. * tests/queryparsertest.cc: Add regression test coverage. Mon Dec 06 01:01:03 GMT 2010 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add coverage for OP_PHRASE as default_op. Sun Dec 05 12:35:32 GMT 2010 Olly Betts <olly@survex.com> * weight/bm25weight.cc: Fix calculation order to avoid inconsistent weights due to rounding when BM25Weight is used with certain non-default parameter combinations. * tests/api_backend.cc: Add regression test bm25weight2. Sun Dec 05 12:10:36 GMT 2010 Olly Betts <olly@survex.com> * backends/chert/chert_compact.cc: Add FIXME comment regarding resolve_duplicate_metadata() getting called multiple times for the same key in multipass mode. Sun Dec 05 12:09:32 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/compactor.h: Add documentation comments. Sun Dec 05 04:54:03 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Omega now wants libmagic-dev (though it's optional currently). Wed Nov 17 10:56:10 GMT 2010 Olly Betts <olly@survex.com> * examples/quest.cc: Add command line options to allow prefixes to be specified for the QueryParser. Tue Nov 09 23:18:36 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc,backends/chert/chert_compact.cc, backends/flint/flint_compact.cc: Fix access to empty priority_queue while merging synonyms (found with _GLIBCXX_DEBUG). Tue Nov 09 03:23:49 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Drop list of platforms valgrind supports, as it takes effort to keep up to date. People can check valgrind.org for an accurate list of currently supported platforms. Sun Nov 07 14:13:44 GMT 2010 Olly Betts <olly@survex.com> * api/compactor.cc,backends/brass/brass_compact.cc, backends/chert/chert_compact.cc,backends/flint/flint_compact.cc, bin/xapian-compact.cc,include/xapian/compactor.h: Compactor::resolve_duplicate_metadata() callback method is now passed an array of std::string plus the array length, which allows for more efficient merging that the series of pairwise merges which was required before. Mon Nov 01 14:47:37 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/multimatch.cc,matcher/multixorpostlist.cc, matcher/multixorpostlist.h: Add debugging to print out the postlist description after recalculation of maxweight, and to MultiXorPostList methods. Fix problem shown by soaktest where the matcher was not told to recalculate the maxweight after a MultiXorPostList child reached end, which was causing an assertion failure in debug builds. (This could also have been causing some performance problems, but I have not measurements.) Mon Nov 01 10:34:30 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/soaktest/soaktest_queries.cc: Fix compilation: call c_str() on argument to atoi, now that util.h no longer defines a string wrapper. Sun Oct 31 10:51:12 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Don't delete this in TermGroup, etc until after we have successfully constructed the Query object to return to avoid a double free in the Query construction throws an exception. Fixes ticket#515. Fri Oct 29 12:46:15 GMT 2010 Olly Betts <olly@survex.com> * tests/api_backend.cc: Fix comment typo in previous commit. Fri Oct 29 12:32:47 GMT 2010 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Empty query now return MSet with firstitem set correctly. * tests/api_backend.cc: Extend msetfirst2 to be a regression test for this issue too. Fri Oct 29 12:13:51 GMT 2010 Olly Betts <olly@survex.com> * api/omenquire.cc: If first is larger than get_doccount() then clamp it to avid pointlessly trying to allocate far too much memory. * tests/api_backend.cc: Regression test msetfirst2. Thu Oct 21 09:38:44 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/Makefile.am: Add zlib-vg.c to distribution tarballs. Wed Oct 20 11:38:13 GMT 2010 Olly Betts <olly@survex.com> * examples/delve.cc: Add '-z' option to count zero-length documents. Mon Oct 18 10:35:46 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Add --quiet/-q option to suppress progress output. Thu Oct 14 03:51:02 GMT 2010 Olly Betts <olly@survex.com> * tests/api_unicode.cc: Fix typo in previous change. Thu Oct 14 01:16:31 GMT 2010 Olly Betts <olly@survex.com> * tests/api_unicode.cc: Expand tested cases to (hopefully) cover all conditional combinations in unicode/utf8itor.cc. Mon Oct 11 11:46:30 GMT 2010 Olly Betts <olly@survex.com> * Makefile.am: Remove xapian-config on "make distclean" rather than "make clean", since configure builds it. Never remove man pages under "make clean". Mon Oct 11 11:41:32 GMT 2010 Olly Betts <olly@survex.com> * tests/Makefile.am: Fix typo - CLEAN_FILES should be CLEANFILES. This means that where zlib-vg.so is used, it now gets cleaned up. Sun Oct 10 11:00:31 GMT 2010 Olly Betts <olly@survex.com> * Makefile.am: Fix so coverage-reconfigure-maintainer-mode adds --enable-maintainer-mode rather than coverage-reconfigure! Sun Oct 10 10:45:38 GMT 2010 Olly Betts <olly@survex.com> * Makefile.am: Factor out COVERAGE_CONFIGURE to make future changes easier. Sun Oct 10 09:59:01 GMT 2010 Olly Betts <olly@survex.com> * Makefile.am: Add coverage-reconfigure-maintainer-mode target which is just like coverage-reconfigure except it also passes --enable-maintainer-mode. Sun Oct 10 07:29:23 GMT 2010 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Check Stem("none") too. Check Stem("").get_description(). No need to check a bogus language name here as stemlangs2 now does that. Sun Oct 10 07:22:41 GMT 2010 Olly Betts <olly@survex.com> * tests/api_stem.cc: Test an invalid languages name with each possible byte value at the start to improve coverage of the switch in api/stem.cc. Sun Oct 10 06:14:42 GMT 2010 Olly Betts <olly@survex.com> * unicode/utf8itor.cc: Correct comments and use bad_cont() in another place (no change to code once inlining is taken into account). Sun Oct 10 06:11:05 GMT 2010 Olly Betts <olly@survex.com> * tests/api_unicode.cc: Improve test coverage for Utf8Iterator::calculate_sequence_length(). Sun Oct 10 00:33:49 GMT 2010 Olly Betts <olly@survex.com> * tests/api_unicode.cc: Tweak testcase utf8iterator2 to add coverage for Utf8Iterator(const char *). Fri Oct 08 13:10:08 GMT 2010 Olly Betts <olly@survex.com> * matcher/selectpostlist.cc,matcher/selectpostlist.h: Implement SelectPostList::check() so that check() on OP_NEAR and OP_PHRASE subqueries won't end up checking potentially huge numbers of documents. Fri Oct 08 12:29:50 GMT 2010 Olly Betts <olly@survex.com> * tests/api_backend.cc: Add testcase phrase3 to provide coverage for SelectPostList::skip_to(). Fri Oct 08 10:36:18 GMT 2010 Olly Betts <olly@survex.com> * matcher/orpostlist.cc,matcher/orpostlist.h: Fix performance regression in some cases caused by the introduction of OrPostList::check(). Thu Oct 07 03:53:11 GMT 2010 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Clear tout before each iteration in scaleweight1. Wed Oct 06 14:30:34 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update from 1.0.22 and ChangeLog. Wed Oct 06 12:39:36 GMT 2010 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add two more testcases for having a phrase generator between prefix and term. Wed Oct 06 12:37:32 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/matchspy.h,include/xapian/postingsource.h: Remove "experimental" marker from PostingSource and ValueCountMatchSpy. Wed Oct 06 12:31:01 GMT 2010 Olly Betts <olly@survex.com> * docs/index.html: Add links to replication and facets documents, and fix typo in serialistion document link. Wed Oct 06 12:12:51 GMT 2010 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/categorisation.rst,docs/facets.rst: Change the categorisation document to talk about facets, since that's the terminology that seems to be most widely used these days, and "categorisation" can also mean automatically assigning categories to documents. Fix up references to features which were removed or changed during development. Wed Oct 06 12:11:21 GMT 2010 Olly Betts <olly@survex.com> * docs/internals.html: Add link to replication protocol. Wed Oct 06 08:36:50 GMT 2010 Olly Betts <olly@survex.com> * examples/simplesearch.cc: Fix cut and paste error in usage message. Wed Oct 06 00:12:40 GMT 2010 Olly Betts <olly@survex.com> * api/emptypostlist.cc,matcher/queryoptimiser.cc: In the query optimiser, use value range bounds to check for value ranges which must be empty. * tests/api_opvalue.cc: Add testcase valuerange5 to check this optimisation actually fires. Tue Oct 05 03:43:02 GMT 2010 Olly Betts <olly@survex.com> * examples/simplesearch.cc: Fix cut-and-paste error - --version now reports simplesearch not simpleexpand. Tue Oct 05 03:41:48 GMT 2010 Olly Betts <olly@survex.com> * docs/categorisation.rst: Update to use the current ValueCountMatchSpy API. Fri Oct 01 09:53:50 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Simplify defaulting enable_documentation a little. Fri Oct 01 09:29:03 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Default enable_sse to yes to fix build on x86. Fri Oct 01 09:10:34 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Fix typo in SSE handling code which stopped it defaulting as intended. Fri Oct 01 02:00:49 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Fri Oct 01 01:54:41 GMT 2010 Olly Betts <olly@survex.com> * backends/Makefile.mk,backends/brass/brass_compact.cc, backends/byte_length_strings.h,backends/chert/chert_compact.cc, backends/flint/flint_compact.cc: Factor out 3 copies of ByteLengthPrefixedStringItor and ByteLengthPrefixedStringItorGt into their own file. Fri Oct 01 01:28:57 GMT 2010 Olly Betts <olly@survex.com> * backends/Makefile.mk,backends/brass/brass_compact.cc, backends/brass/brass_spelling.cc,backends/chert/chert_compact.cc, backends/chert/chert_spelling.cc,backends/flint/flint_compact.cc, backends/flint/flint_spelling.cc, backends/prefix_compressed_strings.h: Factor out 6 copies of PrefixCompressedStringItor and PrefixCompressedStringWriter and 3 copies of PrefixCompressedStringItorGt into their own file. Thu Sep 30 15:28:49 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Thu Sep 30 15:26:52 GMT 2010 Olly Betts <olly@survex.com> * api/compactor.cc: Need utils.h for stat with a std::string argument. Thu Sep 30 15:23:50 GMT 2010 Olly Betts <olly@survex.com> * tests/api_compact.cc: Remove unwanted check left over from cut and pasting code from a previous testcase. Thu Sep 30 15:21:02 GMT 2010 Olly Betts <olly@survex.com> * tests/api_compact.cc: Convert compaction tests to use the new API. Thu Sep 30 14:54:27 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_compact.cc,backends/chert/chert_compact.cc, backends/flint/flint_compact.cc: Correct filenames in @file doxygen comments. Thu Sep 30 14:46:02 GMT 2010 Olly Betts <olly@survex.com> * api/compactor.cc: Use the ChertVersion, etc classes to make sure that the new database has a new UUID rather than creating a "donor" database and then stealing its version/uuid file. Thu Sep 30 14:33:46 GMT 2010 Olly Betts <olly@survex.com> * api/Makefile.mk,api/compactor.cc,backends/brass/Makefile.mk, backends/brass/brass_compact.cc,backends/brass/brass_compact.h, backends/chert/Makefile.mk,backends/chert/chert_compact.cc, backends/chert/chert_compact.h,backends/flint/Makefile.mk, backends/flint/flint_compact.cc,backends/flint/flint_compact.h,bin/, include/Makefile.mk,include/xapian.h,include/xapian/compactor.h: Convert compaction code into a Xapian::Compactor class, and make xapian-compact a simple wrapper around this new class. (ticket#175) Thu Sep 30 06:16:11 GMT 2010 Olly Betts <olly@survex.com> * bin/: Eliminate uses of <iostream> in bin/xapian-compact-*.cc. Thu Sep 30 05:44:06 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-compact-brass.cc,bin/xapian-compact-chert.cc, bin/xapian-compact-flint.cc: Fix indentation of table data arrays. Thu Sep 30 05:32:44 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-compact-brass.cc,bin/xapian-compact-chert.cc: Only skip producing an output table when there are only some inputs for the termlist - for spellings and synonyms, we want to produce an output in this case. * tests/api_compact.cc: Add regression test compactmissingtables1. Wed Sep 29 11:13:18 GMT 2010 Olly Betts <olly@survex.com> * common/document.h: Initialise docid to 0 when creating a document from scratch, as documented. * tests/api_none.cc: Add regression test document2. * include/xapian/document.h: Document that return value is unreliable in this case prior to the next 1.0 and 1.2 releases. Wed Sep 29 07:06:10 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: FreeBSD and OpenBSD don't need explicit dependency libraries, so set link_all_deplibs_CXX=no there. Wed Sep 29 07:04:20 GMT 2010 Olly Betts <olly@survex.com> * xapian-config.in: Just check @link_all_deplibs_CXX@ which we adjust in configure rather than duplicating configure's list of platforms where explicit dependencies aren't required. Mon Sep 27 04:28:46 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Convert error messages to stdout/stderr to exceptions, in preparation for turning this into an API class. Mon Sep 27 03:50:54 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Add support for compacting to a stub database, which can be one of the inputs (for atomic update). * tests/api_compact.cc: Add testcases compactstub3 and compactstub4 as feature tests for this. Mon Sep 27 03:50:00 GMT 2010 Olly Betts <olly@survex.com> * tests/api_compact.cc: Suppress output from xapian-compact in compactstub2 (which I'd disabled for debugging). Sun Sep 26 13:59:20 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Extend to work on stub database files too. * tests/api_compact.cc: Add feature test compactstub2. Sun Sep 26 13:28:45 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Inputs can now be stub database directories, in which case the databases in the stub are used as inputs. * tests/api_compact.cc: Add feature test compactstub1. Sun Sep 26 11:11:49 GMT 2010 Olly Betts <olly@survex.com> * xapian-config.in: Add --static option which makes other options report values for static linking. Tue Sep 21 10:43:17 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Allow phrase generators between a probabilistic prefix and the term itself. Thu Sep 09 11:18:47 GMT 2010 Olly Betts <olly@survex.com> * INSTALL: Raise recommended GCC version from 3.3 to 4.1, since that's the oldest we regularly test with. Thu Sep 09 11:14:03 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Debian etch was discontinued 6 months ago, so it's no longer useful to document packages for developing Xapian on it. Thu Sep 09 03:15:00 GMT 2010 Olly Betts <olly@survex.com> * docs/deprecation.rst: Actually write down our guidelines for supporting other software. Thu Sep 09 00:20:16 GMT 2010 Olly Betts <olly@survex.com> * docs/replication.rst: Adjust text to reflect conclusions about the issues in ticket#434. Thu Sep 09 00:10:03 GMT 2010 Olly Betts <olly@survex.com> * docs/replication.rst: Don't suggest using a symlink to switch between databases - a stub database is much better. Thu Sep 09 00:07:02 GMT 2010 Olly Betts <olly@survex.com> * docs/replication.rst: Assume xapian tools are installed on PATH not in the current directory for example commands. Thu Sep 09 00:05:22 GMT 2010 Olly Betts <olly@survex.com> * docs/replication.rst: Fix a typo. Mon Sep 06 07:10:02 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Add support for --enable-sse=sse and --enable-sse=sse2 to allow control of which SSE instructions to use. Fri Sep 03 13:08:53 GMT 2010 Richard Boulton <richard@tartarus.org> * bin/xapian-replicate.cc: If a fullcopy was attempted, but was not put live, display an explanatory message (if verbose is true). Fri Sep 03 05:54:09 GMT 2010 Olly Betts <olly@survex.com> * backends/chert/chert_modifiedpostlist.h, backends/flint/flint_modifiedpostlist.h,tests/: Fixes required to build on OpenBSD 4.5 with GCC 3.3.5. Tue Aug 31 14:37:44 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Document cases where FLAG_WILDCARD and FLAG_PARTIAL aren't currently supported. Mon Aug 30 06:27:46 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Actaully make autoconf 2.64 a hard minimum requirement (as HACKING already documented). * HACKING,NEWS,configure.ac: autoconf 2.67 is now used to bootstrap. Mon Aug 30 06:26:03 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Mon Aug 30 06:04:13 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Enable use of SSE maths on x86 by default with Sun's compiler. Sat Aug 28 16:04:45 GMT 2010 Olly Betts <olly@survex.com> * api/omdatabase.cc,include/xapian/queryparser.h, tests/harness/backendmanager.h: None of the workarounds we have which are conditional on __SUNPRO_CC are needed with Sun C++ version 5.8, so only enable them for older versions (which will mean we can drop these workarounds with confidence once older versions are dropped by Sun/Oracle, or once we can find out that they have been - currently all I've managed to discover is that version 5.0 reached "end of service life" in 6/2006). Sat Aug 28 15:33:14 GMT 2010 Olly Betts <olly@survex.com> * configure.ac,tests/api_replicate.cc: Solaris < 10 doesn't have setenv() so we have to use putenv() there. And the value we set XAPIAN_MAX_CHANGESETS to is always constant currently, so set it using constant strings generated by a macro. Sat Aug 28 12:35:25 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Beef up the test for whether -lm is required and add a special case to force it to be for Sun's C++ compiler - there's some interaction with libtool and/or shared objects which means that the configure test doesn't think -lm is needed here when it is. Sat Aug 28 10:59:11 GMT 2010 Olly Betts <olly@survex.com> * api/matchspy.cc,common/const_database_wrapper.cc,net/tcpclient.cc, tests/api_replicate.cc,tests/harness/backendmanager_remotetcp.cc, tests/harness/testsuite.cc,tests/perftest/runprocess.cc, unicode/utf8itor.cc: Fix to compile with Sun C++. Sat Aug 28 10:30:08 GMT 2010 Olly Betts <olly@survex.com> * HACKING,PLATFORMS: Move PLATFORMS information to the wiki and replace with a pointer. Sat Aug 28 03:15:47 GMT 2010 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Need to avoid excess precision on m68k when targeting models 68010, 68020, 68030 as well as 68000. Tue Aug 24 05:54:27 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Update for 1.2.3. Tue Aug 24 05:52:47 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Mon Aug 16 16:46:19 GMT 2010 Olly Betts <olly@survex.com> * common/closefrom.cc: Use /dev/fd on Mac OS X. Mon Aug 16 16:18:26 GMT 2010 Olly Betts <olly@survex.com> * common/closefrom.cc: Need safeerrno.h and safeunistd.h on non-Linux platforms too. Mon Aug 16 15:47:48 GMT 2010 Olly Betts <olly@survex.com> * backends/flint_lock.cc,common/Makefile.mk,common/closefrom.cc, common/closefrom.h,configure.ac,net/progclient.cc: Use closefrom() if available, otherwise provide our own implementation (optimised to some extent for many platforms). Sun Aug 15 12:43:04 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS: Update. Sun Aug 15 12:11:26 GMT 2010 Olly Betts <olly@survex.com> * xapian-core.spec.in: Update BuildRequires to specify libuuid-devel instead of e2fsprogs-devel. Sun Aug 15 11:30:05 GMT 2010 Olly Betts <olly@survex.com> * HACKING,INSTALL,common/safeuuid.h,common/win32_uuid.cc, common/win32_uuid.h,configure.ac: libuuid moved from e2fsprogs to util-linux-ng about a year ago, so update documentation, comments, and configure error messages to reflect this. Issue reported by David Jeske on xapian-devel list. Sun Aug 15 07:58:51 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc: Under __WIN32__, we need <io.h> for _open_osfhandle() (ticket#495). Sun Aug 15 07:10:25 GMT 2010 Olly Betts <olly@survex.com> * common/realtime.h: MSVC doesn't cope with a prototype at function scope with a global namespace qualification (ticket#495). Wed Aug 04 10:44:08 GMT 2010 Olly Betts <olly@survex.com> * backends/flint_lock.cc,backends/multi/multi_alltermslist.cc, bin/xapian-chert-update.cc,queryparser/termgenerator.cc, queryparser/termgenerator_internal.cc: Fix more incorrect @file directives. Wed Aug 04 09:19:10 GMT 2010 Olly Betts <olly@survex.com> * matcher/localsubmatch.h,queryparser/termgenerator_internal.h: Fix incorrect @file doxygen directives. Tue Aug 03 14:42:56 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/fdtracker.h: Fix filename in @file. Tue Aug 03 14:32:45 GMT 2010 Olly Betts <olly@survex.com> * common/contiguousalldocspostlist.h,common/unaligned.h: Fix include guards to match header filename (cosmetic issues only). Mon Aug 02 12:38:59 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/brass/brass_values.cc, backends/chert/chert_table.cc,backends/chert/chert_values.cc, backends/flint/flint_alldocspostlist.cc, backends/remote/remote-document.cc,matcher/multimatch.cc, matcher/remotesubmatch.cc,net/progclient.cc: Use new Literal() feature in debug logging of function calls. Mon Aug 02 12:08:42 GMT 2010 Olly Betts <olly@survex.com> * common/pretty.h: Add Literal() class to allow bypassing the pretty printer for interspersing literal strings. Sun Aug 01 10:19:38 GMT 2010 Olly Betts <olly@survex.com> * api/omdatabase.cc: Database::get_spelling_suggestion() will now suggest a correction even if the passed word is in the dictionary, provided the correction has at least the same frequency. Partly addresses #225. * queryparser/queryparser.lemony: Check spelling even if term is in the database. * docs/spelling.rst: Update to reflect these changes. * tests/api_spelling.cc,tests/queryparsertest.cc: Add test coverage for these changes. Tue Jul 27 15:24:56 GMT 2010 Tim Brody <tdb2@ecs.soton.ac.uk> * xapian-core.spec.in: Add xapian-metadata and cmake related files to RPM packaging. Thu Jul 22 07:13:12 GMT 2010 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Wrap line. Thu Jul 22 07:11:23 GMT 2010 Olly Betts <olly@survex.com> * common/remoteserver.h,net/remoteserver.cc: Pass 1.0 for the end time when relaying a NetworkTimeoutError, not RealTime::now() - any time in the past will do, we just want the operation to time out if it would block. Removed related FIXME which is already resolved. Wed Jul 21 18:00:08 GMT 2010 Olly Betts <olly@survex.com> * unicode/tclUniData.cc: Fix comment - this is Unicode 5.2 data, not 5.1. Thu Jul 15 13:45:05 GMT 2010 Olly Betts <olly@survex.com> * common/realtime.h:Explicitly specify global namespace for xapian_sleep_milliseconds prototype as MSVC seems to need this. Thu Jul 15 13:03:13 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix handling of groups of terms which are all stopwords - in situations where this causes a problem we now disable stopword checks for such groups. (ticket#245) * tests/queryparsertest.cc: Add regression testcases. Thu Jul 15 08:13:03 GMT 2010 Olly Betts <olly@survex.com> * common/fileutils.cc: Fix reversed memcmp() test. Sun Jul 11 14:59:06 GMT 2010 Olly Betts <olly@survex.com> * common/fileutils.cc: Factor out UNCW path check into a helper function. Sun Jul 11 14:56:27 GMT 2010 Olly Betts <olly@survex.com> * common/fileutils.cc,tests/Makefile.am,tests/unittest.cc: Move the tests of resolve_relative_path() into a new "unittest" program so they actually are run by "make check". (ticket#243) Sun Jul 11 13:14:24 GMT 2010 Olly Betts <olly@survex.com> * common/fileutils.cc: Support the \\?\ path syntax. Sat Jul 10 15:49:08 GMT 2010 Olly Betts <olly@survex.com> * common/fileutils.cc,common/fileutils.h: Fix resolve_relative_path() to handle UNC paths. Sat Jul 10 15:27:48 GMT 2010 Olly Betts <olly@survex.com> * common/fileutils.cc: -D__WIN32__ on the g++ command line works for testing on Linux, so no need to have that in the code. Add a couple of UNC path test cases which pass, and a commented-out one which fails. Sat Jul 10 14:58:35 GMT 2010 Olly Betts <olly@survex.com> * common/fileutils.cc,common/fileutils.h: We use these routines to resolve a relative path (in a stub database file) in terms of a second path (the filename of that file), so just instead a routine to do exactly that, which is easier to code and more efficient. The new implementation fixes several bugs with Microsoft Windows paths. Some testcases (currently not used) are now present in fileutils.cc (ticket#243). * backends/dbfactory.cc: Use the new API. Fri Jul 09 07:42:31 GMT 2010 Olly Betts <olly@survex.com> * tests/perftest/perftest.cc: I missed a use of atoi() on std::string but just call .cstr() on the string instead of dragging in utils.h. Fri Jul 09 04:39:40 GMT 2010 Olly Betts <olly@survex.com> * common/utils.cc,common/utils.h: Remove unused std::string to const char * wrappers for standard functions, and remove the only use of rmdir()'s wrapper. Thu Jul 08 15:18:04 GMT 2010 Olly Betts <olly@survex.com> * api/omenquire.cc: Xapian::ESet is a reference counted handle, so it is efficient to return by value - remove FIXME which suggests we should avoid doing so. Thu Jul 08 15:17:20 GMT 2010 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Update FIXMEs - it's too late to fix stuff in 1.1.x! Thu Jul 08 15:13:47 GMT 2010 Olly Betts <olly@survex.com> * api/omenquire.cc: Remove FIXMEs about debug logging which have been addressed by the fairly recent debug logging improvements. Thu Jul 08 15:07:07 GMT 2010 Olly Betts <olly@survex.com> * api/keymaker.cc: Fix comment typos. Thu Jul 08 14:50:59 GMT 2010 Olly Betts <olly@survex.com> * api/replication.cc,backends/brass/,backends/chert/, backends/dbfactory_remote.cc,backends/flint/, backends/remote/remote-database.cc,bin/xapian-progsrv.cc, bin/xapian-tcpsrv.cc,common/,net/,tests/perftest/perftest.cc, tests/perftest/perftest.h: Replace use of OmTime with a double holding a count in seconds since the epoch. Thu Jul 08 12:35:40 GMT 2010 Olly Betts <olly@survex.com> * net/tcpclient.cc: Retry select() if it fails with EINTR while waiting for connect(), and discriminate cases with same failure message to aid debugging. Thu Jul 08 05:09:18 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Remove comment left over from the "boolean exclusive" work. Wed Jul 07 13:18:17 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/types.h: Fix documentation comment for Xapian::timeout type - it holds a time interval in milliseconds not microseconds (the API docs for the methods which use it explicitly document this correctly). Tue Jul 06 15:17:09 GMT 2010 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Fix formatting. Tue Jul 06 14:13:03 GMT 2010 Olly Betts <olly@survex.com> * tests/perftest/perftest.cc: Use str(OmTime::as_double()) rather than trying to assemble a string of a floating point number from strings of the sec and usec values. Tue Jul 06 14:03:30 GMT 2010 Olly Betts <olly@survex.com> * common/debuglog.h,common/remote-database.h: Remove unused '#include "omtime.h"'. * backends/remote/remote-database.cc: Add explicit '#include "omtime.h"'. Mon Jul 05 11:40:42 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Restore iterator to start of where we tried to parse a range if we decide it might be a filter term instead. Clear the error if we decide it is a filter. * tests/queryparsertest.cc: Fix expected test output from qp_value_range4 testcase. Mon Jul 05 06:46:16 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix to be smarter about handling a boolean filter term containing ".." in the presence of valuerangeprocessors. * tests/queryparsertest.cc: Add regression test qp_value_range4. Mon Jul 05 04:10:08 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,queryparser/queryparser.lt: Clean up how value ranges are handled to do the checking of the range in the lexer, which then passes a single token (RANGE) to the parser (instead of a RANGE_START token which is always followed by a RANGE_END token). Fri Jul 02 12:32:37 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Don't pass -mtune=generic unless GCC >= 4.2 is in use (ticket#492). Wed Jun 30 10:54:15 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_postlist.cc: Remove unnecessary NULL check. Identified by Coverity's Scan. * backends/chert/chert_postlist.cc: Same change for chert. Tue Jun 29 12:17:16 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_chunkedlisttable.h, backends/chert/chert_chunkedlisttable.h: Remove unused files. Sun Jun 27 04:31:06 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update for 1.2.2. Sun Jun 27 04:22:28 GMT 2010 Olly Betts <olly@survex.com> * examples/delve.cc: Show the database's UUID. Sun Jun 27 03:30:16 GMT 2010 Olly Betts <olly@survex.com> * NEWS.SKELETON: Add "tools" section. Sun Jun 27 03:03:57 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Update for 1.2.2. Sat Jun 26 15:55:45 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Create from ChangeLog. Sat Jun 26 15:36:37 GMT 2010 Olly Betts <olly@survex.com> * NEWS.SKELETON: Add template for NEWS entry. Sat Jun 26 11:09:03 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc: Sync the table right after the base file, which allows more time for the table changes to be written, and doing the sync together may be more efficient with some Linux kernel versions. * backends/chert/chert_table.cc: Same change for chert. Sat Jun 26 06:46:39 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Reorder the release checklist. Thu Jun 24 08:03:18 GMT 2010 Olly Betts <olly@survex.com> bin/xapian-check.cc: Don't try to check doclengths are consistent between the postlist and termlist tables if it would use more than 1GB of memory, and handle std::bad_alloc or std::length_error. This issue affects sup users, as sup allocates docids such that they are sparse and large docids can easily occur. Wed Jun 23 15:38:25 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/weight.h: Revert Xapian::Weight's copy ctor back to protected as GCC 4.1 fails to compile subclasses when it is private (this appears to be a compiler bug). Wed Jun 23 07:31:19 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS: Thank people who reported bugs fixed in 1.2.1. Wed Jun 23 06:52:54 GMT 2010 Olly Betts <olly@survex.com> * xapian-core.spec.in: Update for 1.2.x - add e2fsprogs-devel to BuildRequires and add new files. Wed Jun 23 05:00:34 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Wording tweak. Wed Jun 23 04:11:30 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Need <cstdio> for sprintf(). Fixes compilation error for some platforms and/or compilers (my guess would be GCC 4.5 - it builds OK with GCC 4.4). (ticket#489) Tue Jun 22 14:45:37 GMT 2010 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update for 1.2.1. Mon Jun 21 16:00:22 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Sync with 1.0.21 and update from ChangeLog. Mon Jun 21 15:15:50 GMT 2010 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Unix putenv() requires the buffer passed to remain valid, as it stores it directly in the environment. It's unclear from MSDN if _putenv() does the same, but _putenv_s() must allocate a copy since it needs to have an '=' in, so use that instead. It also has an interface much like setenv(), so the code looks more similar. Mon Jun 21 06:07:42 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.cc: Add missing const. Mon Jun 21 05:21:32 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.cc: Update to match header ABI fix. Mon Jun 21 03:52:41 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: QueryParser::add_boolean_prefix() now uses two overloaded forms instead of a default parameter so that we don't break the ABI. Fri Jun 18 16:56:59 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS: Add 1.0.21 bug reporters. Fri Jun 18 05:52:00 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.cc, queryparser/queryparser.lemony,queryparser/queryparser_internal.h: Add new optional parameter to QueryParser::add_boolean_prefix() to allow the user to indicate a prefix isn't "exclusive" and that OP_AND should be used to combine multiple instances. Fixes ticket#402. This change should also improve efficiency as it avoids copying the lists of prefixes and compares them more efficiently. Fri Jun 18 02:51:11 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Fix doccomment typo. Thu Jun 17 14:04:35 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/weight.h: Xapian::Weight's copy ctor is now private rather than protected. This is an API and ABI compatible change since there's no definition so a subclass which tried to access it before would fail to link. Thu Jun 17 13:00:47 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Missing changes from previous commit. Thu Jun 17 11:37:30 GMT 2010 Olly Betts <olly@survex.com> * api/valuerangeproc.cc: Add support for open-ended ranges (ticket#480). * docs/queryparser.html,docs/valueranges.rst: Document. * tests/queryparsertest.cc: Add feature tests. Thu Jun 17 06:15:24 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Switch out of IN_GROUP mode when we emit a WILDCARD. Fixeds bug#484. * tests/queryparsertest.cc: Add regression testcase to wildcard1. Thu Jun 17 02:42:16 GMT 2010 Olly Betts <olly@survex.com> * tests/api_backend.cc: Rename failedadd1 to failedreplace1 as it's the replace which fails really. Add failedreplace2 which adds explicit coverage for the case of a failed replace when the database isn't empty to start with. Tue Jun 15 12:45:36 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Note in the description of set_database() what the database is used for. Tue Jun 15 12:26:12 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Fix formatting of bullet list. Mon Jun 14 17:02:10 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/chert/chert_table.cc, backends/flint/flint_table.cc: cancel() now marks the Btree as unmodified. * tests/api_backend.cc: Add regression test failedadd1. Sun Jun 13 16:07:49 GMT 2010 Olly Betts <olly@survex.com> * common/Makefile.mk,common/safeuuid.h,configure.ac: Fix mingw build to use __WIN32__ UUID API. Sun Jun 13 12:25:24 GMT 2010 Olly Betts <olly@survex.com> * tests/perftest/freemem.cc: Fix whitespace to match Omega's version. Sun Jun 13 11:52:24 GMT 2010 Olly Betts <olly@survex.com> * tests/perftest/freemem.cc: Merge in changes from omega's version. Clean up whitespace issues. Put the __WIN32__ case last, as we tend to elsewhere (since most people reading the code probably want to see the other version). Sun Jun 13 11:32:30 GMT 2010 Olly Betts <olly@survex.com> * tests/perftest/perftest.cc: "MSWin32" isn't the OS name. Also, prefer += to build up a string. Sat Jun 12 09:31:28 GMT 2010 Olly Betts <olly@survex.com> * tests/api_percentages.cc: Note that 1.0.x returned 4% for the top hit in topercent5, so checking it gets at least 50% is a regression test for that issue. Fri Jun 11 15:38:26 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: -march=pentium4 doesn't seem to give a measurable speed-up or size reduction (from Richard's tests) and it carries a small risk of introducing instructions which don't work on some obscure CPU which implements SSE2, so drop it. Fri Jun 11 15:25:16 GMT 2010 Olly Betts <olly@survex.com> * weight/bm25weight.cc: Stop forcing the wdf_max value to be at least one in BM25Weight::get_maxpart() - this is no longer needed now we calculate percentages based on the number of matching subqueries, and it is more natural for a non-existent term to get zero weight (ditto for a term which always has wdf = 0. * tests/api_anydb.cc: Change qterminfo1 to check that a non-existent term gets zero weight (previously we checked that it got a non-zero weight, but really we want to ensure that it contributes to percentage calculations, which new test topercent5 now does). Fri Jun 11 15:19:29 GMT 2010 Olly Betts <olly@survex.com> * tests/api_percentages.cc: Add topercent5 testcase which checks that an OR search including an non-existent term doesn't get 100%, and that the non-existent term doesn't score more than terms which do exist. Fri Jun 11 14:15:07 GMT 2010 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: When using SSE FP instructions on x86 we can disable the use of volatile since there are no excess precision issues to work around. Fri Jun 11 08:21:04 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Put SSE flags in AM_CXXFLAGS not CXXFLAGS. Fri Jun 11 07:08:01 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Default to building with SSE FP instructions for x86. This avoids issues with excess precision and it a bit faster too. Should fix ticket#487. Thu Jun 10 15:15:33 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Thu Jun 10 15:00:12 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Capitalise "Fedora". Thu Jun 10 11:27:18 GMT 2010 Olly Betts <olly@survex.com> * INSTALL: Reword UUID section - list platforms where we make use of built-in APIs first, give the URL for e2fsprogs home page, and give the package name for Debian and Ubuntu. Thu Jun 10 01:08:21 GMT 2010 Olly Betts <olly@survex.com> * HACKING: libtool 2.2.10 used for snapshots and releases now. Thu Jun 10 00:21:49 GMT 2010 Olly Betts <olly@survex.com> * configure.ac,include/xapian/version_h.cc, include/xapian/visibility.h: We do need to have the visibility annotations on for code compiling against the library, so substitute the probed value of XAPIAN_ENABLE_VISIBILITY into <xapian/version.h> so it is available. Wed Jun 09 11:31:14 GMT 2010 Olly Betts <olly@survex.com> * INSTALL,configure.ac,include/xapian/visibility.h: Control use of GCC's visibility support with defined(XAPIAN_ENABLE_VISIBILITY) instead of !defined(XAPIAN_DISABLE_VISIBILITY), so it is only used when building the library, not when building code which uses it. Add check to configure that GCC actually supports visibility for the platform being built for, which fixes compiler warnings with platforms which don't (such as Mac OS X and mingw). Wed Jun 09 11:17:58 GMT 2010 Olly Betts <olly@survex.com> * HACKING: "private", etc as "access specifiers", not "visibility specifiers". Wed Jun 09 11:04:55 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Note flex and bison needed to build doxygen. Wed Jun 09 05:25:10 GMT 2010 Olly Betts <olly@survex.com> * common/win32_uuid.cc: Fix signed/unsigned comparison warning on mingw. Tue Jun 08 16:56:26 GMT 2010 Olly Betts <olly@survex.com> * common/safewinsock2.h: Reword #error to avoid single quote which gives compilation warning with some GCC 4.2.1 on mingw. Tue Jun 08 14:32:28 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_check.h,backends/chert/chert_check.h, backends/flint/flint_check.h: Remove mutable from std::ostream & member - mutable doesn't make sense for a reference. Tue Jun 08 13:03:56 GMT 2010 Olly Betts <olly@survex.com> * common/,matcher/,weight/weightinternal.cc: Replace the still rather contorted mechanism for accumulating rel termfreqs for local databases with a simpler mechanism which is more efficient in both space and time. Mon Jun 07 16:31:24 GMT 2010 Olly Betts <olly@survex.com> * api/maptermlist.h,api/matchspy.cc,api/termlist.cc,backends/brass/, backends/chert/,backends/flint/, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/multi/multi_termlist.cc,backends/multi/multi_termlist.h, backends/remote/net_termlist.cc,backends/remote/net_termlist.h, common/ortermlist.h,common/termlist.h,expand/ortermlist.cc: Remove default OrTermList::skip_to() implementation and implement skip_to() in each class instead, as this allows a slightly more efficient implementation, and also avoids problems with calling skip_to() as the first operation for the many subclasses which have a "started" flag, which OrTermList::skip_to() won't set. Mon Jun 07 06:08:53 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Rename TermList class to Terms to avoid confusion with the TermList typedef used elsewhere. Add documentation comments for Terms and TermGroup classes. Mon Jun 07 02:18:05 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/: Switch to using io_utils instead of brass_io. Mon Jun 07 02:16:30 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS: Add Arvid Ephraim Picciani for help fixing bootstrap for BusyBox sha1sum. Mon Jun 07 01:16:57 GMT 2010 Olly Betts <olly@survex.com> * common/io_utils.cc: Whitespace tweak. Mon Jun 07 00:58:26 GMT 2010 Olly Betts <olly@survex.com> * common/io_utils.h: Fix header guard macro name to follow convention. Sun Jun 06 15:07:55 GMT 2010 Olly Betts <olly@survex.com> * HACKING,configure.ac: Upgrade to using libtool 2.2.8. Sun Jun 06 04:34:13 GMT 2010 Olly Betts <olly@survex.com> * bin/Makefile.mk,configure.ac: Fix build failures with some combinations of backends disabled (partially addresses ticket#361). Sat Jun 05 16:30:57 GMT 2010 Olly Betts <olly@survex.com> * matcher/remotesubmatch.cc: Add missing parameter of RemoteSubMatch::start_match() to debug logging. Sat Jun 05 15:03:56 GMT 2010 Olly Betts <olly@survex.com> * matcher/queryoptimiser.h: Make QueryOptimiser::optimise_query()'s argument const pointer, since we only pass it to do_subquery() which takes a const pointer. Fri Jun 04 16:05:01 GMT 2010 Olly Betts <olly@survex.com> * common/rset.h: Just keep a reference to the set<docid> from the Xapian::RSet object rather than copying it (since we now only create the RSetI object transiently, so the Xapian::RSet object is definitely valid for its whole lifetime). Fri Jun 04 15:57:50 GMT 2010 Olly Betts <olly@survex.com> * api/omenquire.cc,matcher/localmatch.cc,matcher/localmatch.h: Keep a Xapian::RSet handle in the LocalSubMatch object and only create an RSetI object briefly in LocalSubMatch::prepare_match(). Fri Jun 04 15:36:45 GMT 2010 Olly Betts <olly@survex.com> * common/pretty.h: RSetI objects are no longer passed as parameters or return values, so we no longer need to handle them. Fri Jun 04 15:28:21 GMT 2010 Olly Betts <olly@survex.com> * api/omenquire.cc,common/esetinternal.h,common/omenquireinternal.h, common/rset.h,expand/esetinternal.cc,matcher/rset.cc: RSetI is used when collating statistics prior to performing the match, and when generating an ESet. The second use is entirely unnecessary and just ends up with us copying the std::set<Xapian::docid> from inside the Xapian::RSet object, so eliminate it and just use the Xapian::RSet object directly. This allows RSetI to be simplified somewhat too. Fri Jun 04 14:22:38 GMT 2010 Olly Betts <olly@survex.com> * tests/api_backend.cc: Make arrays static const. Fri Jun 04 13:25:20 GMT 2010 Olly Betts <olly@survex.com> * backends/Makefile.mk: Note need to update include/xapian/version_h.cc when adding a new backend. * AUTHORS: Thanks to Conrad Hoffmann for pointing this omission out. Thu Jun 03 15:01:44 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/soaktest/soaktest_queries.cc: Add missing include of str.h Thu Jun 03 12:09:35 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/Makefile.am: Fix building of zlib-vg.so in VPATH builds. Thu Jun 03 11:49:31 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/orpostlist.cc,tests/api_backend.cc: Fix several bugs in OrPostList::check() (introduced since 1.2.0). Fixes ticket #485. Specifically, we have to check a sub-postlist even if head=did, because we need to know if that sub-postlist is valid. Also, if a sub-postlist is not valid, set its head to did+1, since we know that we're no longer interested in matches before did, and this allows the OrPostList to assume that the minimum docid of its children is the current docid. Thu Jun 03 07:15:24 GMT 2010 Olly Betts <olly@survex.com> * matcher/: Implement MultiXorPostList which returns correct weights for XOR of more than 2 subqueries, and drop XorPostList completely. (ticket#475) * tests/api_anydb.cc: Add regression test xor2. Thu Jun 03 04:38:56 GMT 2010 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Tweak loop end test to be a normal while. Thu Jun 03 01:22:20 GMT 2010 Olly Betts <olly@survex.com> * tests/api_stem.cc: Note stem2's feature was backported to 1.0.21. Tue Jun 01 13:49:44 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc,backends/brass/brass_termlist.cc, backends/chert/chert_database.cc,backends/chert/chert_termlist.cc, backends/flint/flint_database.cc,backends/flint/flint_document.cc, backends/flint/flint_termlist.cc,matcher/: Fix LOGCALL_VOID uses for ctors and dtors to be LOGCALL_CTOR and LOGCALL_DTOR respectively. Tue Jun 01 13:33:07 GMT 2010 Olly Betts <olly@survex.com> * api/omdocument.cc: Use LOGCALL_VOID instead of LOGCALL with a void return type. Tue Jun 01 13:08:02 GMT 2010 Olly Betts <olly@survex.com> * matcher/multimatch.cc: The debug logging now copes with being passed RSet* so remove special casing with ?:. Tue Jun 01 10:17:27 GMT 2010 Olly Betts <olly@survex.com> * HACKING,configure.ac: Drop support for --enable-log=profile - dedicated profiling tools are likely to return more useful results. * common/debuglog.cc,common/debuglog.h: Rename xapian_debuglogger__ global to xapian_debuglogger_ to avoid reserved __ in the name. * common/omdebug.cc,common/omdebug.h: Remove in favour of debuglog.h. * common/pretty.h: New PrettyOStream<> template class for pretty printing types in debug logs (and eventually the testsuite). * api/,backends/brass/,backends/chert/, backends/dbfactory_remote.cc,backends/flint/, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_document.cc, backends/inmemory/inmemory_positionlist.cc, backends/multi/multi_postlist.cc,backends/multi/multi_termlist.cc, common/,expand/esetinternal.cc,expand/expandweight.cc, expand/ortermlist.cc,matcher/,net/progclient.cc, net/remoteconnection.cc: Migrate all remaining uses of omdebug.h to debuglog.h. Write more parameters to the debug log. Retire the barely used APICALL category and use API instead (reflecting actual current use). Add new REPLICA category. Fix names logged for ctors and dtors which have "::" in the class name. Sun May 30 13:45:12 GMT 2010 Olly Betts <olly@survex.com> * api/omenquire.cc: Remove bogus "Data::" from debug logging for two methods. Sun May 30 08:49:22 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/stem.h,languages/stem.cc: Xapian::Stem now recognises "nb" and "nn" as codes for the Norwegian stemmer. * tests/api_stem.cc: Add feature test stem2 to check that these new codes work. Sat May 29 10:12:40 GMT 2010 Olly Betts <olly@survex.com> * common/debuglog.h: Add overload of operator| which writes ", " and then a value to std::ostream. * api/omdatabase.cc,api/replication.cc, backends/brass/brass_alldocspostlist.cc, backends/brass/brass_metadata.cc,backends/brass/brass_table.cc, backends/chert/,backends/dbfactory.cc, backends/flint/flint_alldocspostlist.cc,expand/esetinternal.cc, expand/expandweight.cc,matcher/,net/remoteconnection.cc, weight/bm25weight.cc,weight/weight.cc: Use this in LOGCALL uses to make them shorter and more readable. Fri May 28 06:47:57 GMT 2010 Olly Betts <olly@survex.com> * api/leafpostlist.cc,api/omdatabase.cc,backends/brass/brass_table.cc, backends/chert/chert_table.cc,backends/dbfactory.cc, backends/flint/flint_alldocspostlist.cc: Put LOGCALL macro use all on one line so we can more easily mechanise use and update. Fri May 28 04:22:35 GMT 2010 Olly Betts <olly@survex.com> * api/replication.cc,api/valueiterator.cc, backends/brass/brass_alltermslist.cc, backends/chert/chert_alltermslist.cc,backends/dbfactory.cc, backends/inmemory/inmemory_document.cc, backends/remote/remote-document.cc,matcher/synonympostlist.cc, net/remoteconnection.cc,weight/bm25weight.cc: Use NO_ARGS instead of "" and LOGCALL_VOID instead of LOGCALL with void return type. Fri May 28 04:12:35 GMT 2010 Olly Betts <olly@survex.com> * matcher/valuerangepostlist.cc: Add missing inclusions of "debuglog.h". * backends/brass/brass_metadata.h,backends/brass/brass_postlist.h, backends/brass/brass_synonym.h,backends/chert/chert_metadata.h, backends/chert/chert_postlist.h,backends/chert/chert_synonym.h, backends/dbfactory_remote.cc,backends/flint/flint_metadata.h, backends/flint/flint_synonym.h,common/database.h,common/rset.h: Remove unused inclusions of "omassert.h" and "omdebug.h". * common/rset.h: Remove unused inclusion of "weightinternal.h". * api/omdatabase.cc,api/registry.cc,api/replication.cc,backends/brass/, backends/chert/,backends/dbfactory.cc,backends/flint/: Add missing explicit inclusions of "debuglog.h". Fix LOGCALL markup of ctors and dtors. Use LOGCALL instead of DEBUGCALL. Fri May 28 04:06:08 GMT 2010 Olly Betts <olly@survex.com> * backends/flint/flint_alldocspostlist.h: Add missing explicit inclusions of "autoptr.h", "flint_database.h", and forward declaration of class FlintCursor. Fri May 28 03:52:40 GMT 2010 Olly Betts <olly@survex.com> * common/const_database_wrapper.h,common/databasereplicator.h, common/gnu_getopt.h: Add "(C)" to copyright statements for consistency with other files. Fri May 28 03:45:26 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Include "debuglog.h" instead of "omdebug.h". Fri May 28 03:45:03 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.cc: Remove unused inclusion of "omdebug.h". Fri May 28 03:40:50 GMT 2010 Olly Betts <olly@survex.com> * expand/esetinternal.cc,matcher/localmatch.cc,matcher/multimatch.cc, matcher/rset.cc: Add missing explicit inclusions of "omassert.h". Thu May 27 23:39:57 GMT 2010 Olly Betts <olly@survex.com> * common/Makefile.mk,common/expand.h: Remove unused header. Thu May 27 06:45:27 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/stem.h: Update the list of languages that the Xapian::Stem constructor recognises. Thu May 27 05:20:42 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/stem.h,languages/Makefile.mk, languages/compiler/generator.c,languages/stem.cc, languages/steminternal.cc,languages/steminternal.h,tests/Makefile.am, tests/api_stem.cc: Add the ability to specify your own stemming algorithm by subclassing Xapian::StemImplementation, mostly based on patch from Evgeny Sizikov in ticket#448. Thu May 27 04:50:31 GMT 2010 Olly Betts <olly@survex.com> * tests/api_opvalue.cc: Use TEST_REL() instead of tout and TEST() with an inequality. Thu May 27 04:28:07 GMT 2010 Olly Betts <olly@survex.com> * api/replication.cc,backends/databasereplicator.cc,common/debuglog.h, common/omdebug.h,expand/esetinternal.cc,expand/ortermlist.cc, matcher/,net/remoteconnection.cc: Partial mass conversion of files from omdebug.h to debuglog.h (this takes care off all the files without BSN copyright). Thu May 27 04:24:36 GMT 2010 Olly Betts <olly@survex.com> * common/output.h: Use <ostream> not <fstream>. Thu May 27 04:09:25 GMT 2010 Olly Betts <olly@survex.com> * matcher/orpostlist.cc: Prefer ++x to x += 1 for consistency. Wed May 26 13:32:20 GMT 2010 Olly Betts <olly@survex.com> * matcher/exactphrasepostlist.cc: Separate (C) notice from @file doccomment. Wed May 26 13:11:57 GMT 2010 Olly Betts <olly@survex.com> * backends/flint/flint_postlist.cc,backends/flint/flint_postlist.h: Move virtual method FlintPostList::get_doclength() definition out of header, which also means the header no longer needs omdebug.h. Wed May 26 13:03:22 GMT 2010 Olly Betts <olly@survex.com> * common/debuglog.h: Strip out commented out RETURN_VOID machinery. This case is automatically handled provided LOGCALL_VOID is used. Wed May 26 12:33:04 GMT 2010 Olly Betts <olly@survex.com> * common/debuglog.h: Don't check the category more than once in ctors and dtors. Don't convert the return value to a string in RETURN() unless we want to log it. Wed May 26 02:33:00 GMT 2010 Olly Betts <olly@survex.com> * tests/api_opvalue.cc: Convert file description to doccomment. Remove unused headers and reorder remainder more standardly. Tue May 25 10:17:42 GMT 2010 Olly Betts <olly@survex.com> * tests/api_backend.cc: Note ordecay1 and ordecay2 fixes now also in 1.0.21. Whitespace tweak. Tue May 25 04:11:53 GMT 2010 Olly Betts <olly@survex.com> * tests/api_backend.cc: make_ordecay1_db and make_ordecay2_db end up looping over at least 4 billion values (or possibly infinitely) on platforms where sizeof(long) == 8. The two databases are very similar and the latter one works for the former test, so use the same db for both tests, and rewrite the generator to work portably. Sun May 23 07:30:58 GMT 2010 Olly Betts <olly@survex.com> * tests/api_backend.cc: Note XOR decay bug also fixed in 1.0.21. Sat May 22 13:31:38 GMT 2010 Olly Betts <olly@survex.com> * configure.ac,tests/Makefile.am: Only try to build zlib-vg.so if we are found a suitable valgrind installed (otherwise it's a waste of effort to build it, and the build will fail without the valgrind headers). Sat May 22 13:18:33 GMT 2010 Olly Betts <olly@survex.com> * configure.ac,tests/Makefile.am,tests/runtest.in,tests/zlib-vg.c: zlib can produce "uninitialised" output from "initialised" input - the output does decode to the input, so this is presumably just some unused bits in the output, so we use an LD_PRELOAD hack to get valgrind to check the input is initialised and then tell it that the output is initialised. Wed May 19 14:34:52 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/andmaybepostlist.cc,matcher/andmaybepostlist.h, matcher/orpostlist.cc: Fix a bug causing documents to be skipped when OR decayed to AND_MAYBE, if the new RHS was behind the LHS at the time of the decay. Fixes #476. * tests/api_backend.cc: Add test case ordecay2 as a regression test of this. Also, remove some debug output in ordecay1, to speed it up quite a bit. Wed May 19 11:48:07 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/orpostlist.cc,tests/api_backend.cc: Fix a bug in the decay from OR to AND: if the decay happened when the two sub-postlists were on different docids, but the higher of those docids actually occurred in both postlists, the higher docid would be incorrectly skipped. Add regression test ordecay1 to test this. Wed May 19 10:36:55 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_backend.cc,tests/api_compact.cc, tests/harness/testrunner.cc,tests/harness/testrunner.h: Add new test backend selection flag "generated", to select only backends which support generated test databases. Wed May 19 03:14:15 GMT 2010 Olly Betts <olly@survex.com> * api/,backends/brass/,backends/chert/, backends/contiguousalldocspostlist.cc,backends/flint/, backends/inmemory/inmemory_database.cc, backends/remote/remote-database.cc,backends/slowvaluelist.cc, common/utils.h,expand/esetinternal.cc,matcher/phrasepostlist.cc, net/remoteserver.cc,net/remotetcpclient.cc, queryparser/termgenerator.cc,tests/, tests/harness/backendmanager_multi.cc, tests/harness/backendmanager_remote.cc, tests/harness/backendmanager_remotetcp.cc,tests/internaltest.cc, tests/perftest/perftest.cc,tests/perftest/perftest_matchdecider.cc, tests/perftest/perftest_randomidx.cc,tests/queryparsertest.cc, tests/termgentest.cc: Fully migrate from om_tostring() to str(). Wed May 19 03:13:43 GMT 2010 Olly Betts <olly@survex.com> * cmake/xapian-config-version.cmake,cmake/xapian-config.cmake: Remove generated files from SVN. Wed May 19 02:59:47 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-inspect.cc: Add misisng explicit '#include <utils.h>'. Tue May 18 14:24:03 GMT 2010 Olly Betts <olly@survex.com> * bin/Makefile.mk,bin/xapian-check.cc,bin/xapian-compact.cc: Improve building with flint, chert and/or brass disabled (partially addresses ticket#361). Tue May 18 00:42:08 GMT 2010 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix warnings with --disable-backend-remote (partially addresses ticket#361). Sat May 15 06:56:26 GMT 2010 Olly Betts <olly@survex.com> * net/progclient.cc: Read the max fd with sysconf() instead of using a hardcoded value of 256. Work even if stdin and stdout have been closed. Thu May 13 10:53:24 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/orpostlist.cc: Correct some debugging lines. Thu May 13 09:57:12 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/xorpostlist.cc,matcher/xorpostlist.h: Pass a minweight value of 0 to children of XOR postlists, because we need to know about all matching documents from the children even if they have a low weight. * tests/api_backend.cc: Regression test for this: xordecay1. Thu May 13 08:47:22 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/soaktest/soaktest_queries.cc: Add more code comments, increase the number of repetitions from 100 to 10000, remove accidentally added semicolons after some braces, remove "return false" at end of test which was added for debugging purposes, clear tout before testing each query, and display the query number, the query, and the msets in the event of failure. Tue May 11 10:16:05 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/soaktest/soaktest_queries.cc: Don't fail for empty result sets. Tue May 11 10:13:08 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/soaktest/: Add a "soaktest" testsuite, intended to contain long-running tests with random data. Currently contains a single test which builds and runs random queries, checking that the results returned are consistent when asking for different result ranges. Mon May 10 16:57:29 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_opvalue.cc: Add tests for OP_VALUE_RANGE to get coverage of the skip_to() and check() methods of ValueRangePostList. Mon May 10 16:51:14 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_opvalue.cc: Add a test to get coverage of ValueGePostList::skip_to(). Sun May 09 21:22:50 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_nodb.cc,tests/api_opvalue.cc: Move test "opvaluege1" from api_nodb.cc to api_opvalue.cc, and rename to valuege3 for consistency with other tests of OP_VALUE_GE. Sun May 09 21:19:43 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/Makefile.am,tests/api_anydb.cc,tests/api_opvalue.cc: Move the tests for OP_VALUE_* query operators from api_anydb.cc to api_opvalue.cc. Remove copyright statements earlier than 2007, since OP_VALUE_* wasn't added until 2007. Sun May 09 21:04:06 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/valuegepostlist.cc: Initialise valuelist in check() if necessary. * tests/api_anydb.cc: Add regression test for this. Thu May 06 15:03:43 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_replicate.cc: Add some extra debugging, to help identify the problem with ticket #472. Thu May 06 13:38:25 GMT 2010 Olly Betts <olly@survex.com> * common/safeuuid.h,configure.ac,INSTALL: Make use of built-in UUID API on FreeBSD and NetBSD. (ticket#470) Thu May 06 13:10:30 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS,Makefile.am,cmake/,configure.ac: Install files to make Xapian easier to use with cmake. Thanks to Sune Vuorela for providing the desired output. Thu May 06 11:42:16 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/fdtracker.cc: Don't pass NULL to closedir(). Thu May 06 11:19:41 GMT 2010 Olly Betts <olly@survex.com * tests/api_anydb.cc: Test UUID is 36 characters long. Thu May 06 10:41:21 GMT 2010 Olly Betts <olly@survex.com> * matcher/valuegepostlist.cc: Use "str.h" and str(). Thu May 06 10:01:48 GMT 2010 Olly Betts <olly@survex.com> * matcher/valuegepostlist.cc: Remove unused headers. Thu May 06 09:55:27 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/valuegepostlist.cc,matcher/valuerangepostlist.cc: Keep return value from valuelist->get_value() as a const string & rather than a string - could be more efficient on some compilers, though testing with gcc 4.1 revealed no difference. Thu May 06 09:49:12 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/valuegepostlist.cc: "while(" -> "while (" in recent patch. Thu May 06 09:30:27 GMT 2010 Olly Betts <olly@survex.com> * docs/replication.rst: Remove experimental warning. Thu May 06 05:05:57 GMT 2010 Olly Betts <olly@survex.com> * matcher/valuerangepostlist.cc: Use "str.h" and str(). Thu May 06 05:03:22 GMT 2010 Olly Betts <olly@survex.com> * matcher/valuerangepostlist.cc: Remove '#include "leafpostlist.h"' and '#include "document.h"' which are no longer used. Thu May 06 04:53:07 GMT 2010 Olly Betts <olly@survex.com> * matcher/valuerangepostlist.cc: Remove '#include "autoptr.h"' which is no longer required. Thu May 06 04:18:25 GMT 2010 Olly Betts <olly@survex.com> * matcher/valuerangepostlist.cc: "while(" -> "while (" in recent patch. Wed May 05 16:44:17 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/orpostlist.cc,matcher/orpostlist.h: Add OrPostList::check() method, which calls check() on the sub-postlists, rather than falling back to skip_to(). This is particularly helpful when a sub-postlist is one for which skip_to() can be significantly more expensive than check(), such as a ValueRangePostList. I've observed a speedup of 10 times for such a subquery with this change. Wed May 05 16:34:30 GMT 2010 Richard Boulton <richard@tartarus.org> * matcher/valuegepostlist.cc,matcher/valuerangepostlist.cc, matcher/valuerangepostlist.h: Implement ValueRangePostList and ValueGePostList using a ValueList streaming values for the value slot of interest, rather than using an AllDocsPostList to get all documents, using that to open temporary documents for each document being considered, and then using that to get the value. I've observed a speedup of more than 70 times for range queries in a real-world situation with this change (a 29 second search became a 0.4 second search). (ticket#432) Wed May 05 06:44:31 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Wed May 05 05:42:56 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_cursor.cc,backends/chert/chert_cursor.cc, backends/flint/flint_cursor.cc: If we have to rebuild the cursor on next() because the table changed underneath us, we need to advance it because find_entry() will put the cursor on the entry we were already on, or the one before if the entry we were on has been removed. Bug uncovered by notmuch testsuite, but I'm unable to reduce it to a sane testcase for our testsuite. (Debian#579951) Tue May 04 19:38:31 GMT 2010 Richard Boulton <richard@tartarus.org> * api/replication.cc,bin/xapian-replicate.cc, common/replicatetcpclient.h,common/replication.h, net/replicatetcpclient.cc,tests/api_replicate.cc: Fix a bug in replication, which triggered when multiple modifications were made to a database while a full copy of the database was in progress. In this situation, the client would fail to apply more than one changeset before trying (and failing) to make the database live, due to an incorrect assignment of the new revision number to "offline_needed_revision" instead of "offline_revision". Also, replication now only sleeps for a period (to allow reader to catch up) after a changeset has been applied to a live database - previously, it would sleep unnecessarily after a changeset had been applied to an offline database. Also, make the period for which replication sleeps in this situation configurable with the "-r" command line option to xapian-replicate. A test for these changes will be committed to xapian-bindings shortly. Sat May 01 15:25:48 GMT 2010 Olly Betts <olly@survex.com> * Merge change accidentally committed to brass-btree branch: * tests/perftest/runprocess.cc: Use safesyswait.h, fixing build failure on "make check" on FreeBSD. Fri Apr 30 12:33:25 GMT 2010 Richard Boulton <richard@tartarus.org> * backends/brass/brass_databasereplicator.cc, backends/chert/chert_databasereplicator.cc, backends/flint/flint_databasereplicator.cc,tests/api_replicate.cc: If a DB file doesn't exist when applying a changeset, try creating it - this happens with lazy tables which aren't created in the initial revision copied. Add replicate4 test to check that this works correctly with a lazily created position table. Fixes ticket #468. Fri Apr 30 07:09:31 GMT 2010 Olly Betts <olly@survex.com> * docs/deprecation.rst: Replication and PostingSource aren't experimental in 1.2.x. Fri Apr 30 00:52:22 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Fix typo in previous commit. Thu Apr 29 23:46:09 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Check is SA_SIGINFO is defined before using it as it isn't available everywhere. Fixes FTBFS on GNU Hurd. Thu Apr 29 12:11:48 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Install headers under PREFIX/include not PREFIX/include/xapian. This broke compiling code which didn't use XO_LIB_XAPIAN or xapian-config. Wed Apr 28 12:05:23 GMT 2010 Olly Betts <olly@survex.com> * Makefile.am,m4-macros/xapian-1.1.m4: Remove -1.1 suffix from xapian-1.1.m4 and libxapian-1.1.la. Wed Apr 28 06:15:45 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Sync with 1.0.20. * NEWS,configure.ac: 1.2.0! Tue Apr 27 12:45:29 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS: Add bug reporter. Mon Apr 26 08:26:26 GMT 2010 Olly Betts <olly@survex.com> * common/getopt.cc,common/gnu_getopt.h: On Cygwin, #include <getopt.h> for declarations of optind, etc rather than providing our own declarations. Should fix warnings in Cygwin build reported by Reini Urban on xapian-devel list. Mon Apr 26 06:27:07 GMT 2010 Olly Betts <olly@survex.com> * common/getopt.cc,common/gnu_getopt.h: Revert previous 2 getopt-related patches. The latest state was breaking on Solaris which (contrary to what its man pages say) prototypes getopt() stuff in stdlib.h. This seems to be a rabbit hole we don't want to head further down. Sat Apr 24 02:12:53 GMT 2010 Olly Betts <olly@survex.com> * examples/delve.cc: Update --help output for -v to cover all terms case. Sat Apr 24 01:58:00 GMT 2010 Olly Betts <olly@survex.com> * examples/delve.cc: Add -a option to list all terms in a database. Sat Apr 24 01:51:01 GMT 2010 Olly Betts <olly@survex.com> * examples/copydatabase.cc: Prefer string() to "" converted to string. Thu Apr 22 14:21:30 GMT 2010 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix bug related to ticket#464 (the previous fix doesn't fix some similar cases). * tests/api_backend.cc: Add regression test msize2. Thu Apr 22 11:02:11 GMT 2010 Olly Betts <olly@survex.com> * tests/: Hook up --abort-on-error fix for termgentest, stemtest, queryparsertest, and internaltest. Thu Apr 22 06:06:53 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Add a hyphen. Wed Apr 21 11:16:50 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/testrunner.cc: Exit with status 1 not 0 if we caught an exception from the harness itself. Wed Apr 21 11:11:10 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/testrunner.cc,tests/harness/testsuite.cc: Fix --abort-on-error to actually work. Tue Apr 20 12:53:29 GMT 2010 Olly Betts <olly@survex.com> * docs/deprecation.rst: Note how to disable deprecation warnings. Tue Apr 20 07:27:23 GMT 2010 Olly Betts <olly@survex.com> * api/omenquire.cc,matcher/multimatch.cc: Fix incorrect estimate and bounds on matches when sorting and collapsing in a particular case (ticket#464). * tests/api_backend.cc: Add regression test msize1. Mon Apr 19 01:59:59 GMT 2010 Olly Betts <olly@survex.com> * examples/delve.cc: Use strtoul() instead of atoi() to decode document id and value slot numbers passed on the command line. Report out of range and invalid values. Sat Apr 17 09:17:25 GMT 2010 Olly Betts <olly@survex.com> * common/getopt.cc,common/gnu_getopt.h: Fix getopt code to work on platforms without GNU getopt but with getopt(). Fri Apr 16 14:23:49 GMT 2010 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update for 1.1.5. * NEWS: Sync changes from 1.0.19. Fri Apr 16 12:29:14 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Handle NULL in the const char * catch. Fri Apr 16 05:47:43 GMT 2010 Olly Betts <olly@survex.com> * api/postingsource.cc: Fix wrapping. Thu Apr 15 09:07:40 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_replicate.cc: Use a fixed size buffer for copying a file, rather than a dynamically sized buffer; avoids using the GCC extension allowing dynamically sized arrays. Thu Apr 15 04:02:34 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS: Update with bug reporters. Thu Apr 15 00:30:45 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/termgenerator.h: Note how TermGenerator handles stopwords. Wed Apr 14 23:41:55 GMT 2010 Olly Betts <olly@survex.com> * common/getopt.cc,common/gnu_getopt.h: Put our optarg, etc in a namespace and then pull them in to avoid linking to versions in the C library. Fixes warnings on Cygwin. Wed Apr 14 15:33:39 GMT 2010 Olly Betts <olly@survex.com> * tests/api_backend.cc: Remove already fixed FIXME comment. Wed Apr 14 14:13:10 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Use const reference to avoid a needless handle copy. Wed Apr 14 14:00:21 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_alltermslist.cc, backends/brass/brass_alltermslist.h, backends/chert/chert_alltermslist.cc, backends/chert/chert_alltermslist.h, backends/flint/flint_alltermslist.cc, backends/flint/flint_alltermslist.h: Fix memory leak if Database throws an exception in XXXAllTermsList constructor. Fixes bug#462. * tests/api_backend.cc: Enable qpmemoryleak1 for non-remote backends as a regression test for this fix. Tue Apr 13 12:38:08 GMT 2010 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Don't memcpy() a block to itself - it's a waste of effort, and (probably) undefined behaviour (as a block overlaps itself). Tue Apr 13 12:36:53 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/chert/chert_table.cc: Don't memcpy() a block to itself - it's a waste of effort, and (probably) undefined behaviour (as a block overlaps itself). Tue Apr 13 08:38:40 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Add a ParserHandler class to take care of releasing the yyParser object. Tue Apr 13 07:49:33 GMT 2010 Olly Betts <olly@survex.com> * common/autoptr.h: Fix comment typo. Tue Apr 13 07:47:40 GMT 2010 Olly Betts <olly@survex.com> * api/replication.cc: "autoptr.h" not <autoptr.h> (it's an in-tree header). Tue Apr 13 07:25:11 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix leak of yyParser if an exception is thrown during parsing. Partly fixes ticket#462. * tests/api_backend.cc: Add regression test qpmemoryleak1. Tue Apr 13 01:37:08 GMT 2010 Olly Betts <olly@survex.com> * common/utils.cc,common/utils.h: Revert change to make removedir() recurse into subdirectories - it doesn't fix a bug, so isn't appropriate when we're trying to get 1.2.0 out (backends don't currently put files in subdirectories, which was the justification in the commit message). Mon Apr 12 11:50:37 GMT 2010 Olly Betts <olly@survex.com> * Backport change from brass: * backends/flint/: Fix to rebuild cursors if the table gains or loses a level. Fixes ticket#455. Mon Apr 12 09:29:39 GMT 2010 Olly Betts <olly@survex.com> * Backport change from brass: * backends/chert/: Fix to rebuild cursors if the table gains or loses a level. Fixes ticket#455. Mon Apr 12 09:26:53 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/: Fix to rebuild cursors if the table gains or loses a level. Fixes ticket#455. Mon Apr 12 08:27:29 GMT 2010 Olly Betts <olly@survex.com> * tests/api_none.cc: Work around compilation failure on OpenBSD. Fixes ticket#458. Fri Apr 02 04:03:04 GMT 2010 Olly Betts <olly@survex.com> * m4-macros/xapian-1.1.m4: Give a more specific error message for the cases where XAPIAN_CONFIG isn't found, is a directory, or isn't executable. Thu Mar 25 14:43:57 GMT 2010 Olly Betts <olly@survex.com> * tests/api_backend.cc: Add databasemodified1 testcase to provide some test coverage for DatabaseModifiedError. Thu Mar 25 06:32:44 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-check-flint.cc: Report docid for doclength mismatch. Thu Mar 25 06:23:26 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-check-brass.cc,bin/xapian-check-chert.cc: Report docid for doclength mismatch. Sat Mar 20 15:04:18 GMT 2010 Olly Betts <olly@survex.com> * docs/spelling.rst: Add a note about removing automatically added spelling dictionary entries. Sat Mar 20 09:30:44 GMT 2010 Olly Betts <olly@survex.com> * docs/spelling.rst: Add section on indexing. Move the "algorithm" section to the end, as it is really just background information for the curious. Wed Mar 17 01:33:39 GMT 2010 Olly Betts <olly@survex.com> * INSTALL: Add a hyphen. Wed Mar 17 01:30:34 GMT 2010 Olly Betts <olly@survex.com> * INSTALL: Correct description of --enable-assertions. It does NOT enable debugging symbols, and shouldn't control checks on bad data passed to API calls (if it does anywhere, that's a bug). Note that Xapian will run more slowly with assertions on. Wed Mar 17 01:10:13 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_values.cc: Put ValueUpdater in Brass namespace to avoid cross-linking issues with ValueUpdater class in chert (change taken from brass-btree branch). Sun Mar 14 10:05:36 GMT 2010 Olly Betts <olly@survex.com> * Backport change from chert: * backends/flint/flint_database.cc, backends/flint/flint_positionlist.cc, backends/flint/flint_positionlist.h: When updating a document, rather than decoding the old positions, comparing with the new, and then encoding the new if different, we now just encode the new and then compare the encoded forms. (ticket#428) This change also avoids calling FlintPositionListTable::del() when we know the entry doesn't already exist. * backends/flint/flint_positionlist.cc: Minor refactor to improve code clarity and to only create std::string variable when it is needed. Sun Mar 14 09:45:34 GMT 2010 Olly Betts <olly@survex.com> * Backport change from brass: * backends/chert/chert_positionlist.cc: Minor refactor to improve code clarity and to only create std::string variable when it is needed. Sun Mar 14 09:29:54 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_positionlist.cc: Minor refactor to improve code clarity and to only create std::string variable when it is needed. Sun Mar 14 07:37:03 GMT 2010 Olly Betts <olly@survex.com> * api/replication.cc,tests/api_closedb.cc,tests/api_metadata.cc, tests/internaltest.cc: Catch by const reference rather than just reference, unless we actually need to modify the caught object. Sun Mar 14 06:45:11 GMT 2010 Olly Betts <olly@survex.com> * api/replication.cc: Check live_db only has a single sub-database even if we just lazily opened it. Sun Mar 14 05:00:21 GMT 2010 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc, backends/flint/flint_databasereplicator.cc: Add missing explicit '#include "str.h"'. Sat Mar 13 09:22:24 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_replicate.cc: Always use off_t to refer to the size of a file: fixes warnings about comparting a signed value to an unsigned value. Thu Mar 11 16:07:10 GMT 2010 Richard Boulton <richard@tartarus.org> * api/replication.cc: Open the livedb lazily if it gets left in a closed state (due to an exception when applying a changeset), rather than reopening it straight away in an exception handler. This requires the livedb to become a mutable member, but is safer in the event that an exception is thrown by the WritableDatabase constructor. Thu Mar 11 15:17:12 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_replicate.cc: Uncomment the rmtmpdir() at the end of replicate3; it was accidentally committed debugging code. Thu Mar 11 13:03:57 GMT 2010 Richard Boulton <richard@tartarus.org> * api/replication.cc: If an exception is thrown, mark that the offline revision needed is unknown (represented by being empty); ensures that a database isn't put live incorrectly in this case. Also, actually read the message for REPL_REPLY_END_OF_CHANGES, so that we get an error if the message is truncated. Also, if an attempt to apply a changeset to the live database fails with an exception while the live database is closed, reopen the live database before allowing the exception to propagate. * tests/api_replicate.cc: Split replicate() helper function into two parts, to separate getting the changeset messages, and applying them. Add replicate3 test, which checks that replications which are provided with truncated messages raise appropriate errors. Thu Mar 11 00:46:39 GMT 2010 Olly Betts <olly@survex.com> * common/replicate_utils.cc: Avoid creating temporary string objects to reduce bloat in compiled code. Remove explicit std:: qualifiers. Thu Mar 11 00:38:00 GMT 2010 Olly Betts <olly@survex.com> * common/replicate_utils.cc: The directory we're writing changesets to will always exist, so no need to ensure it does. Thu Mar 11 00:10:19 GMT 2010 Olly Betts <olly@survex.com> * backends/chert/chert_databasereplicator.cc, backends/flint/flint_databasereplicator.cc: Fix calculation of end pointer for changes file. Since the merge, it has been off by CONST_STRLEN(CHANGES_MAGIC_STRING). Wed Mar 10 23:59:38 GMT 2010 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc, backends/chert/chert_databasereplicator.cc: Add missing explicit '#include "str.h"'. Wed Mar 10 13:07:41 GMT 2010 Richard Boulton <richard@tartarus.org> * common/Makefile.mk,common/replicate_utils.cc, common/replicate_utils.h: Add some utility functions for replication to create changeset files, and append content to them, to avoid duplicating the code in each backend. * backends/chert/chert_database.cc, backends/chert/chert_databasereplicator.cc, backends/chert/chert_databasereplicator.h, backends/flint/flint_database.cc, backends/flint/flint_databasereplicator.cc, backends/flint/flint_databasereplicator.h: Make the DatabaseReplicator subclasses respect the XAPIAN_MAX_CHANGESETS environment variable, such that they write the changesets they're applying to disk if the value is non-zero. Enables replication from a replicated copy. * tests/api_replicate.cc: Refactor some common code, and add replicate2, which tests replication from a replicated copy. Skip this test for brass, since brass doesn't support this. Wed Mar 10 12:11:41 GMT 2010 Richard Boulton <richard@tartarus.org> * common/utils.cc,common/utils.h: Make removedir() recurse into subdirectories; this ensures that replication works if backends put some files in subdirectories of their replicas (eg, if the replica is itself a replicated database). Wed Mar 10 12:07:40 GMT 2010 Richard Boulton <richard@tartarus.org> * backends/chert/,backends/flint/,common/Makefile.mk, common/io_utils.cc,common/io_utils.h: Refactor duplicated code from {flint,chert}_io.{cc,h} into common/io_utils.{cc,h}. Rename flint_io_read() to io_read(); do similar for flint_io_write() and flint_io_sync(), and for the chert versions. Leave brass alone to avoid conflicting with the new brass backend development branch. Mon Mar 08 02:32:35 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-check-chert.cc: Check that the doclen initial chunk exists. Tue Mar 02 13:28:50 GMT 2010 Olly Betts <olly@survex.com> * Backport change from brass: * backends/chert/chert_database.cc, backends/chert/chert_positionlist.cc, backends/chert/chert_positionlist.h: When updating a document, rather than decoding the old positions, comparing with the new, and then encoding the new if different, we now just encode the new and then compare the encoded forms. (ticket#428) This change also avoids calling ChertPositionListTable::del() when we know the entry doesn't already exist. Tue Mar 02 04:00:22 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Harness now catches const char * exceptions explicitly so it can show their contents. Tue Mar 02 03:59:27 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/valueiterator.h: Comment tweak. Tue Mar 02 03:57:55 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Add missing forward declaration of "class Database". Tue Mar 02 01:51:33 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update partially from ChangeLog. Mon Mar 01 05:38:41 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc, backends/brass/brass_positionlist.cc, backends/brass/brass_positionlist.h: When updating a document, rather than decoding the old positions, comparing with the new, and then encoding the new if different, we now just encode the new and then compare the encoded forms. (ticket#428) This change also avoids calling BrassPositionListTable::del() when we know the entry doesn't already exist. Mon Mar 01 02:09:40 GMT 2010 Olly Betts <olly@survex.com> * net/serialise.cc: Don't call Document::values_count() twice when we can trivially avoid doing so. Ditto for termlist_count() and positionlist_count(). Mon Mar 01 01:53:58 GMT 2010 Olly Betts <olly@survex.com> * bin/: Rename tot_off to last_docid when it is actually holding the last_docid value. Sun Feb 28 23:47:49 GMT 2010 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Whitespace tweak. Sun Feb 28 23:44:58 GMT 2010 Olly Betts <olly@survex.com> * tests/dbcheck.cc: Tweak whitespace to match standard usage. Catch exceptions by const reference. * HACKING: Discuss why exceptions should be caught by const reference. Tweak markup of blockquotes. Sun Feb 28 22:50:50 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Add note about "new SomePODType()" not being the same as "new SomePODType". Sun Feb 28 22:29:31 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Quote ``cat`` and ``more``. Sun Feb 28 14:00:39 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_compact.cc: Clarify comment about regression test for 1.1.4, and remove a test line which is implicit in the following dbcheck() line anyway. Sun Feb 28 11:07:33 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/fdtracker.cc,tests/harness/fdtracker.h: Add GPL boilerplate and copyright notices which I failed to do before committing these new files. Thu Feb 25 10:34:15 GMT 2010 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: Initialise z_stream members zalloc, zfree, and opaque with Z_NUL rather than 0 cast to the appropriate type, as that's what the zlib documentation says to do. Add missing initialisation of opaque for the inflate z_stream which the docs say is needed, though reading the zlib code, isn't actually for current versions. Wed Feb 24 14:24:05 GMT 2010 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc: Fix indentation. Wed Feb 24 14:18:25 GMT 2010 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remote-database.h, common/remoteprotocol.h,common/remoteserver.h, docs/remote_protocol.html,net/remoteserver.cc,tests/api_spelling.cc: Add remote backend support for WritableDatabase::add_spelling() and WritableDatabase::remove_spelling(). This bumps the remote protocol to version 35.0 (so both client and servers will need updating). Suggesting spelling corrections isn't yet supported. (ticket#178) Wed Feb 24 09:09:27 GMT 2010 Olly Betts <olly@survex.com> * examples/delve.cc: "lastdocid" -> "highest document id ever used" in output so it's clearer what this number means. Tue Feb 23 14:03:47 GMT 2010 Olly Betts <olly@survex.com> * api/postingsource.cc: Fix incorrect classname in error message. Tue Feb 23 13:16:11 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: For -fno-strict-aliasing for GCC 4.2 to avoid bad code being generated due to a bug in that compiler version. Fixes ticket#449. Tue Feb 23 03:08:02 GMT 2010 Olly Betts <olly@survex.com> * configure.ac,tests/harness/testsuite.cc: Use sigsetjmp() and siglongjmp() where available so that the set of blocked signals get restored and the test harness can catch a second incidence of a particular signal in a run. Use sigaction() instead of signal() where available, which allows us to report the address associated with SIGSEGV, SIGFPE, SIGILL, and SIGBUS. Mon Feb 22 14:29:50 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_databasereplicator.cc, backends/chert/chert_databasereplicator.cc, backends/flint/flint_databasereplicator.cc: Fix exception message if we get an unsupported changeset type. Mon Feb 22 01:49:03 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/dbcheck.cc: Check the wdf, doclen, and value bounds reported by the database against correct (tight) bounds calculated by iterating through the database contents. Mon Feb 22 01:47:20 GMT 2010 Richard Boulton <richard@tartarus.org> * backends/brass/brass_database.cc,backends/chert/chert_database.cc: Ensure that wdfs are reported to the stats object correctly when replacing documents. These checks were absent entirely in chert, and incorrectly skipped in some cases in brass. Mon Feb 22 00:59:57 GMT 2010 Olly Betts <olly@survex.com> * tests/api_valuestream.cc: Reset tout before testing each interval. Prefer string() to "". Mon Feb 22 00:57:59 GMT 2010 Olly Betts <olly@survex.com> * docs/deprecation.rst,examples/delve.cc: Remove delve's -k option, accepted as an undocumented alias for -V since 0.9.10 for compatibility with 0.9.9 and earlier. Just use -V instead. Mon Feb 22 00:45:37 GMT 2010 Olly Betts <olly@survex.com> * examples/delve.cc: If documents are specified, change "-V<valueno>" to only show calues for those documents. Sun Feb 21 14:11:58 GMT 2010 Olly Betts <olly@survex.com> * common/stringutils.h: Factor out looking up a character in the table into a separate inlined function. Add C_islcxdigit() to check for an lower case hex digit. Sun Feb 21 11:10:10 GMT 2010 Olly Betts <olly@survex.com> * examples/Makefile.mk,examples/xapian-metadata.cc: Add new example "xapian-metadata" which allows you to get and set individual user metadata entries. Sun Feb 21 10:31:56 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Belatedly update version number to 1.1.4. Sat Feb 20 01:09:45 GMT 2010 Olly Betts <olly@survex.com> * tests/runtest.in: Now we have fd leak checking in the test harness, remove the ulimit as it doesn't add anything, and has caused problems on some Debian buildds. Fri Feb 19 02:08:25 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/backendmanager.h,tests/harness/testsuite.cc: Call backendmanager->clean_up() before checking for fd leaks to avoid finding the sockets/pipes as leaks. This means that clean_up() needs to handle being called more than once for a test sometimes (which the current versions do) so document that. Thu Feb 18 02:31:17 GMT 2010 Olly Betts <olly@survex.com> * examples/simpleindex.cc: Explicitly call commit() on the database, as that is good practice (since you see any exceptions). Thu Feb 18 01:52:10 GMT 2010 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Clear tout on each pass of the loop to avoid a lot of irrelevant output under --verbose if the test fails. Thu Feb 18 01:33:53 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Fix fd tracking change to work when HAVE_VALGRIND isn't defined. Thu Feb 18 00:46:20 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Document the possible exception messages from QueryParser::parse_query(). Thu Feb 18 00:20:18 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Also explicitly commit after indexing for flint and chert. Wed Feb 17 15:58:04 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/: Add machinery to check for leaked FDs. Currently requires /proc/self/fd to work (so it works on at least Linux), but should do nothing on platforms without this directory. Wed Feb 17 15:53:30 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Explicitly commit after indexing for brass, so we see any exceptions from the commit (commit from the dtor swallows any exceptions). Tue Feb 16 16:23:30 GMT 2010 Richard Boulton <richard@tartarus.org> * bin/xapian-compact.cc: Fix failure to set lastdocid to an appropriate value when using --no-renumber. * tests/api_compact.cc: Add regression test for this. Tue Feb 16 16:10:49 GMT 2010 Richard Boulton <richard@tartarus.org> * examples/delve.cc: Display the lastdocid value when displaying general database statistics. Tue Feb 16 15:39:50 GMT 2010 Richard Boulton <richard@tartarus.org> * bin/xapian-check-brass.cc,bin/xapian-check-chert.cc: Correct typo in error message. Mon Feb 15 13:59:45 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Update for 1.1.4. Mon Feb 15 13:58:44 GMT 2010 Olly Betts <olly@survex.com> * PLATFORMS: We don't support GCC < 3.1, so remove discussion of issues with GCC 2.95. Mon Feb 15 13:55:10 GMT 2010 Olly Betts <olly@survex.com> * PLATFORMS: Update from 1.0.18. Mon Feb 15 11:51:15 GMT 2010 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Update from 1.0.18. Sun Feb 14 05:47:30 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix QueryParser to handle an explicit AND before a hated term (foo AND -bar). (ticket#447) * tests/queryparsertest.cc: Add regression test cases. Fri Feb 12 01:27:32 GMT 2010 Olly Betts <olly@survex.com> * queryparser/termgenerator_internal.cc: Only include trailing '+' or '#' on a term if it isn't followed by a word character (makes more sense and matches QueryParser's behaviour). (ticket#446) * tests/termgentest.cc: Add regression test. Thu Feb 11 04:30:56 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-inspect.cc: Add new "open" command to allow easy switching between tables. Wed Feb 10 02:40:36 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/document.h: Rewrite the note about Document being a lazy handle to fix a few details and avoid nailing down the exact implementation. Wed Feb 10 00:37:54 GMT 2010 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Update for chert and other recent changes. Wed Feb 10 00:37:04 GMT 2010 Olly Betts <olly@survex.com> * docs/replication_protocol.rst: Update. Mon Feb 08 12:05:05 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_version.cc: Fix comments about BRASS_VERSION which were previously just cut-and-pasted from those about CHERT_VERSION. Mon Feb 08 11:54:04 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_metadata.cc: Include missing headers. Mon Feb 08 09:23:00 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_document.h,backends/chert/chert_document.h: Fix typos in documentation comments. Mon Feb 08 08:27:09 GMT 2010 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Remove unused headers <streambuf> and "omdebug.h". Mon Feb 08 01:28:48 GMT 2010 Olly Betts <olly@survex.com> * common/debuglog.h: Use std::boolalpha so bool is logged as "true" or "false" (instead of "1" or "0"). Sat Feb 06 10:04:42 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS: Sync with 1.0 branch. Sat Feb 06 08:20:31 GMT 2010 Olly Betts <olly@survex.com> * tests/api_closedb.cc: Avoid trying to create a database where an already open one exists (which fails on Microsoft Windows). (ticket#443). Fri Feb 05 13:28:08 GMT 2010 Olly Betts <olly@survex.com> * examples/simpleexpand.cc,examples/simpleindex.cc, examples/simplesearch.cc: Handle --help and --version. Fri Feb 05 13:25:37 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Fix not to assign elements off the end of a vector (ticket#440). Fri Feb 05 11:41:05 GMT 2010 Richard Boulton <richard@tartarus.org> * api/matchspy.cc,include/xapian/matchspy.h,tests/api_matchspy.cc: Remove ValueCountMatchSpy::get_values(). values_begin() should be used instead to iterate over the values. Fri Feb 05 07:39:46 GMT 2010 Olly Betts <olly@survex.com> * configure.ac: Enable automake's std-options feature to check all installed programs support --help and --version. Fri Feb 05 01:45:49 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/document.h: Add Document::add_boolean_term(). * tests/api_none.cc: Add testcase document1 as a feature test. Fri Feb 05 00:27:52 GMT 2010 Olly Betts <olly@survex.com> * backends/chert/chert_btreebase.cc,backends/flint/flint_btreebase.cc: Add missing SEQUENTIAL entry to documentation of base file format. Thu Feb 04 03:06:26 GMT 2010 Olly Betts <olly@survex.com> * tests/api_compact.cc: Fix to work with MSVC makefiles (ticket#440). Thu Feb 04 03:04:43 GMT 2010 Olly Betts <olly@survex.com> * HACKING,common/Makefile.mk,common/safesyswait.h: Add wrapper for <sys/wait.h> which defines macros for WEXITSTATUS and WIFEXITED when required. Wed Feb 03 15:01:36 GMT 2010 Richard Boulton <richard@tartarus.org> * api/matchspy.cc,include/xapian/matchspy.h,tests/api_matchspy.cc: Add ValueCountMatchSpy::values_begin() and values_end() to allow iteration over all the values in a ValueCountMatchSpy without needing to handle the dict. Move the contents of ValueCountMatchSpy into a reference counted internal object, so that the iterator can ensure that the contents aren't deleted before it finishes. Wed Feb 03 15:01:02 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/generate-api_generated: Add ValueCountMatchSpy to the list of non-copyable API classes. Wed Feb 03 13:18:03 GMT 2010 Richard Boulton <richard@tartarus.org> * include/xapian/matchspy.h,tests/api_matchspy.cc: Provide an iterator for accessing the top values found by a matchspy, rather than using a method which takes a vector reference as a parameter. Should be wrappable much more nicely (and get wrapped automatically), and be easier to use: the testcase for this is significantly shorter than for the old API. Remove StringAndFrequency from API - it's now just an internal class. * api/matchspy.cc: Test new access method for top_values. Wed Feb 03 06:46:37 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS: Add Emmanuel Engelhart for reporting problems indexing Burmese text. Wed Feb 03 06:32:03 GMT 2010 Olly Betts <olly@survex.com> * include/xapian/unicode.h,queryparser/queryparser.lemony, queryparser/termgenerator_internal.cc: Add ENCLOSING_MARK and COMBINING_SPACING_MARK categories to is_wordchar(). Make TermGenerator and QueryParser ignore several zero-width space characters. * tests/queryparsertest.cc,tests/termgentest.cc: Add test coverage for the changes. Wed Feb 03 03:14:30 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix apostrophe abuse in comment. Wed Feb 03 01:23:04 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Make TermList::as_opwindow_query() private. Wed Feb 03 01:19:19 GMT 2010 Olly Betts <olly@survex.com> * common/replication.h: Use "" not <> for including API headers. Wed Feb 03 01:08:58 GMT 2010 Olly Betts <olly@survex.com> * common/replication.h: Remove warnings about this being an experimental API, which no longer make sense now it is internal. Tue Feb 02 21:59:59 GMT 2010 Olly Betts <olly@survex.com> * tests/generate-api_generated: Remove DatabaseReplica. Tue Feb 02 17:01:23 GMT 2010 Richard Boulton <richard@tartarus.org> * include/xapian/document.h: Add a note to the API documentation for the Document class about the potential issues involved in lazy loading. Fixes ticket #421. Tue Feb 02 16:21:39 GMT 2010 Richard Boulton <richard@tartarus.org> * docs/categorisation.rst: Remove documentation of NumericRanges stuff for now, since the NumericRanges is no longer on trunk. Tue Feb 02 15:59:14 GMT 2010 Richard Boulton <richard@tartarus.org> * common/replication.h: Put the ReplicationInfo, DatabaseMaster and DatabaseReplica symbols back into public visibility - they're needed for the tests, and for the command line scripts. Tue Feb 02 15:06:44 GMT 2010 Richard Boulton <richard@tartarus.org> * api/matchspy.cc,include/xapian/matchspy.h,tests/api_matchspy.cc: Remove NumericRanges from trunk - their API is still uncertain, so I want to do a bit more development work on them on the matchspy branch before committing to support them for several years. Tue Feb 02 13:45:24 GMT 2010 Richard Boulton <richard@tartarus.org> * api/replication.cc,backends/brass/brass_database.cc, backends/chert/chert_database.cc,backends/flint/flint_database.cc, common/,include/Makefile.mk,include/xapian/replication.h, net/replicatetcpclient.cc,net/replicatetcpserver.cc, tests/generate-api_generated: Move replication support out of the public API headers and public symbol visibility. Access to replication, for now, is via the command-line scripts: this is similar to the access to running a remote database server, and the access to database compaction and checking. See ticket #347 for some commentary. Tue Feb 02 07:22:28 GMT 2010 Olly Betts <olly@survex.com> * Backport change from chert: * backends/flint/flint_alltermslist.h,backends/flint/flint_spelling.h, backends/flint/flint_spellingwordslist.h: Add missing std:: qualifiers. Tue Feb 02 07:12:51 GMT 2010 Olly Betts <olly@survex.com> * docs/doxygen_api.conf.in: Add generated include/xapian/error.h so that Xapian::Error and subclass API docs are collated for a VPATH build. * docs/Makefile.am: Add regression test to make sure doxygen collated API docs for Xapian::RuntimeError. Tue Feb 02 06:52:55 GMT 2010 Olly Betts <olly@survex.com> * Backport change from brass: * backends/chert/chert_alltermslist.h,backends/chert/chert_spelling.h, backends/chert/chert_spellingwordslist.h: Add missing std:: qualifiers. Tue Feb 02 01:34:52 GMT 2010 Richard Boulton <richard@tartarus.org> * docs/replication.rst: Add note about not using reopen() with databases being updated by the replication client. Tue Feb 02 01:17:34 GMT 2010 Richard Boulton <richard@tartarus.org> * api/matchspy.cc,docs/categorisation.rst,include/xapian/matchspy.h, tests/api_matchspy.cc: Remove score_evenness(), since it turned out to be of little or no use in practical situations. See ticket #435 for more details. Mon Feb 01 21:53:41 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/chert/chert_table.cc: Use LOGCALL_VOID for add() methods which now return void. Mon Feb 01 21:50:05 GMT 2010 Olly Betts <olly@survex.com> * Backport change from brass: * backends/flint/flint_table.cc,backends/flint/flint_table.h: Fix documentation comments for FlintTable::add() and FlintTable::del() - errors actually cause exceptions, not a false return value, and del() returns true if it actually removed an entry. Fix FlintTable::add() to throw an exception for the (hopefully never hit) case when the tag value is too insanely large to handle. With that FlintTable::add() always returns true, and we never check the return value anyway, so just make it return void. Mon Feb 01 21:37:33 GMT 2010 Olly Betts <olly@survex.com> * Backport change from brass: * backends/chert/chert_table.cc,backends/chert/chert_table.h: Fix documentation comments for ChertTable::add() and ChertTable::del() - errors actually cause exceptions, not a false return value, and del() returns true if it actually removed an entry. Fix ChertTable::add() to throw an exception for the (hopefully never hit) case when the tag value is too insanely large to handle. With that ChertTable::add() always returns true, and we never check the return value anyway, so just make it return void. Mon Feb 01 16:02:27 GMT 2010 Richard Boulton <richard@tartarus.org> * bin/xapian-check-brass.cc,bin/xapian-check-chert.cc: Fix checking of the METAINFO key in chert and brass. For small databases, the statistics fit in few enough bytes that the trailing unpack_uint_last() moved to end, so errors weren't being reported, but for large databases the check was incorrectly reporting an error. Mon Feb 01 09:24:14 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_table.cc,backends/brass/brass_table.h: Fix documentation comments for BrassTable::add() and BrassTable::del() - errors actually cause exceptions, not a false return value, and del() returns true if it actually removed an entry. Fix BrassTable::add() to throw an exception for the (hopefully never hit) case when the tag value is too insanely large to handle. With that BrassTable::add() always returns true, and we never check the return value anyway, so just make it return void. Mon Feb 01 01:45:09 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_alltermslist.h,backends/brass/brass_spelling.h, backends/brass/brass_spellingwordslist.h: Add missing std:: qualifiers. Mon Feb 01 00:57:21 GMT 2010 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.h: Add missing std:: qualifier. Mon Feb 01 00:52:27 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_spelling.h,backends/chert/chert_spelling.h: Add missing std:: qualifier. Sun Jan 31 23:35:52 GMT 2010 Olly Betts <olly@survex.com> * backends/chert/chert_termlisttable.cc, backends/chert/chert_termlisttable.h: Remove ChertTermListTable::get_doclength() as it is no longer used. Sun Jan 31 23:19:50 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_termlisttable.cc, backends/brass/brass_termlisttable.h: Remove BrassTermListTable::get_doclength() as it is no longer used. Sun Jan 31 09:19:39 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-inspect.cc: "until" improvements: + No longer reports the current key and tag. + If passed a key before the current one, no longer silently advances to the next key. + Explicitly says if the specified key doesn't exist, including where we ended up. Sun Jan 31 00:19:41 GMT 2010 Olly Betts <olly@survex.com> * tests/api_compact.cc: Use a single Document object rather than creating the same one 10000 times. Sun Jan 31 00:08:38 GMT 2010 Olly Betts <olly@survex.com> * tests/api_compact.cc: Move make_multichunk_db() before the testcase which uses it, rather than between two helper functions for another testcase. Sun Jan 31 00:07:03 GMT 2010 Olly Betts <olly@survex.com> * tests/api_compact.cc: Remove confusing comment, cut-and-pasted from elsewhere and not relevant here. Fri Jan 29 06:13:18 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_metadata.cc,backends/chert/chert_metadata.cc, backends/flint/flint_metadata.cc: Correct filenames in @file. Thu Jan 28 22:34:03 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_compact.cc: Add check that xapian-compact works correctly on a database with multiple chunks for a single term; this is a regression test for ticket #427 Thu Jan 28 22:33:16 GMT 2010 Richard Boulton <richard@tartarus.org> * bin/xapian-compact-brass.cc,bin/xapian-compact-chert.cc: Don't cut the extra byte of doclen chunk keys. Thu Jan 28 21:23:48 GMT 2010 Richard Boulton <richard@tartarus.org> * bin/xapian-compact-flint.cc: Revert previous patch for flint - it only applies to new key format used by chert and brass. Thu Jan 28 21:13:38 GMT 2010 Richard Boulton <richard@tartarus.org> * bin/xapian-compact-brass.cc,bin/xapian-compact-chert.cc, bin/xapian-compact-flint.cc: Partial fix for #427: truncate follow-on keys for postlist chunks fully (was leaving the trailing 0-byte, causing the code for chert and brass to incorrectly think a new chunk had started, resulting in a truncated posting list. Thu Jan 28 20:42:47 GMT 2010 Richard Boulton <richard@tartarus.org> * bin/xapian-inspect.cc: Add "until" command to display keys until a specified one. Useful for seeing all keys for a particular posting list, etc. Thu Jan 28 01:31:54 GMT 2010 Olly Betts <olly@survex.com> * tests/dbcheck.cc: Whitespace tweak. Wed Jan 27 06:29:21 GMT 2010 Olly Betts <olly@survex.com> * tests/dbcheck.cc: <config.h> not "config.h". Wed Jan 27 04:20:42 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_inverter.h: Make two getters const. Wed Jan 27 04:01:17 GMT 2010 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_database.h, backends/flint/flint_database.cc,backends/flint/flint_database.h: Add ChertWritableDatabase::insert_mod_plist() to handle the simpler case where we know the document doesn't already exist, to avoid adding extra code to that common path over the code we had before the #250-related changes. Tue Jan 26 03:25:59 GMT 2010 Olly Betts <olly@survex.com> * tests/internaltest.cc: In testcase pack_uint_preserving_sort1, use data() not c_str(), calculate the end position just once, and use swap to copy the current packed to prev_packed. Fri Jan 15 12:51:38 GMT 2010 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/flint/flint_database.cc: Fix to use 'M' as the flag character for modified entries in modplists (not 'U'). We never explicitly check for 'M' (we check for 'A' vs 'D' vs something else), so this isn't a functional change, but rather making the values consistent (as they were before the fix for #250). Fri Jan 15 08:25:14 GMT 2010 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_database.h, backends/flint/flint_database.cc,backends/flint/flint_database.h: Use Xapian::termcount_diff instead of Xapian::termcount for parameters of add_freq_delta() which can be signed. Fix parameter docs to say "change" instead of "increase". Fri Jan 15 06:05:28 GMT 2010 Olly Betts <olly@survex.com> * tests/api_replacedoc.cc: Just test that begin == end when we expect no positions rather than generating a string representation and then testing if it is empty. Thu Jan 14 11:17:36 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Auto-enabling of -Werror is now done for GCC 4.1 or newer. Thu Jan 14 05:06:42 GMT 2010 Olly Betts <olly@survex.com> * HACKING: Include the MIT licence text in HACKING itself. * COPYING_MIT: Remove external copy of text - it could mislead users into thinking that this is the current licence of Xapian. Sun Jan 10 07:36:16 GMT 2010 Olly Betts <olly@survex.com> * configure.ac:With --enable-maintainer-mode, enable -Werror for GCC >= 4.1 rather than >= 4.0 as Apple's GCC 4.0 gives bogus uninitialised variable warnings for pack.h. Sat Jan 09 00:56:06 GMT 2010 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Add support for quoting boolean terms so they can contain arbitrary characters (partly addresses ticket#128). * tests/queryparsertest.cc: Add test coverage. Fri Jan 08 13:25:25 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_compact.cc: Add copyright note for my changes committed earlier today. Fri Jan 08 12:19:07 GMT 2010 Olly Betts <olly@survex.com> * bin/xapian-compact-brass.cc,bin/xapian-compact-chert.cc: Fix merging of databases by xapian-compact for brass and chert. * tests/api_compact.cc: Enable test compactmerge1 for brass and chert as a regression test. Fri Jan 08 11:26:41 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc,backends/chert/chert_database.cc, backends/flint/flint_database.cc: Rearrange comparison between old and new termlists to avoid double check of termlist.at_end(). Fri Jan 08 11:22:30 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/api_compact.cc: Add test of xapian-compact when merging databases. Currently, the generated databases are broken for chert and brass, so skipping the test for those to avoid breaking the snapshotter. Fri Jan 08 11:13:13 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/: Move routines for helping to check database contents and consistency into dbcheck.cc, with headers in dbcheck.h. Fri Jan 08 10:07:09 GMT 2010 Richard Boulton <richard@tartarus.org> * tests/internaltest.cc: Add test_pack_uint_preserving_sort1() which I wrote when suspicious of that function - passes fine, but probably worth keeping the test coverage. Fri Jan 08 00:23:06 GMT 2010 Olly Betts <olly@survex.com> * AUTHORS: Add Henry Combrinck for Search::Xapian patches. Thu Jan 07 15:51:57 GMT 2010 Richard Boulton <richard@tartarus.org> * api/postingsource.cc: For ValueWeightPostingSources for which there are no entries, the upper bound returned will be an empty string, which translated to -inf when passed to sortable_unserialise(). This was causing the matcher to get confused and return no results in some queries involving such a posting source. This commit fixes this. * tests/api_postingsource.cc: Add emptyvalwtsource1 as a regression test for this, and a general test of this area of code. Thu Jan 07 09:17:58 GMT 2010 Olly Betts <olly@survex.com> * backends/brass/brass_inverter.h: Fix comment typo. Wed Dec 30 11:46:45 GMT 2009 Richard Boulton <richard@tartarus.org> * backends/brass/brass_database.cc,backends/chert/chert_database.cc, backends/flint/flint_database.cc: When updating documents, don't update posting entries which havn't changed. Largely fixes ticket #250. * AUTHORS: Add Kan-Ru Chen, since his patch attached to ticket #250 was the basis of this fix. Wed Dec 30 11:44:18 GMT 2009 Richard Boulton <richard@tartarus.org> * backends/brass/brass_inverter.h: Add methods for updating an existing posting, so we don't have to remove then re-add it. Wed Dec 30 11:41:57 GMT 2009 Richard Boulton <richard@tartarus.org> * backends/brass/brass_database.h: Remove old unused members: total_length and lastdocid. Wed Dec 30 01:07:04 GMT 2009 Richard Boulton <richard@tartarus.org> * backends/chert/chert_database.h: Remove old unused members: total_length and lastdocid. Wed Dec 30 00:21:03 GMT 2009 Richard Boulton <richard@tartarus.org> * backends/chert/chert_database.cc,backends/chert/chert_database.h, backends/flint/flint_database.cc,backends/flint/flint_database.h: Move frequently repeated code for updating the freq_deltas and mod_plists items into separate functions. Makes the code considerably clearer, and prepares the way for applying the patch to fix ticket #250. Wed Dec 30 00:07:33 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_replacedoc.cc: Add a few more sequences of operations which look worth testing, after looking at the code implementing replace_document(). Tue Dec 29 22:29:08 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_replacedoc.cc: Check consistency of the posting lists with the termlists, and consistency of the statistics, at each step of modtermwdf1. Also, add some other terms to the document, so we can check the stored doclength for terms which haven't been modified. Sun Dec 27 14:56:16 GMT 2009 Richard Boulton <richard@tartarus.org> * backends/chert/chert_modifiedpostlist.cc, backends/flint/flint_modifiedpostlist.cc: Report the wdf from the modifications, rather the sum of the wdf from the modifications and the old wdf. The old behaviour was just wrong, but we had no test coverage for this until the last commit. Sun Dec 27 13:34:06 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_replacedoc.cc: Add test of the contents of the postlist of a term which is being modified. Also, tidy up the output of docterms_to_string() to skip any empty position lists. Sat Dec 26 19:21:19 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_replacedoc.cc: Extend test to cover removing a term but not deleting it, and finally deleting the document. Thu Dec 24 19:18:41 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_replacedoc.cc: Expand test to cover some modification to the positional information. Thu Dec 24 19:06:32 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_replacedoc.cc: Add some more sets of modification operations to the tests. Thu Dec 24 18:55:37 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_replacedoc.cc: Add test modtermwdf1, which tests a lot more cases where documents are modified. Thu Dec 24 14:36:38 GMT 2009 Olly Betts <olly@survex.com> * tests/: Canonicalise the conditions on testcases. Thu Dec 24 14:08:59 GMT 2009 Olly Betts <olly@survex.com> * tests/api_closedb.cc: Fix typo in last commit. Thu Dec 24 13:39:28 GMT 2009 Olly Betts <olly@survex.com> * tests/api_closedb.cc: Check WritableDatabase::close() during a transaction does *NOT* implicitly call commit(). Thu Dec 24 13:07:09 GMT 2009 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc,backends/brass/brass_database.h, backends/chert/chert_database.cc,backends/chert/chert_database.h, backends/flint/flint_database.cc,backends/flint/flint_database.h: Fix WritableDatabase::close() to commit() changes (unless a transaction is in progress). * tests/api_closedb.cc: Add regression test closedb4. Wed Dec 23 00:26:34 GMT 2009 Olly Betts <olly@survex.com> * backends/brass/brass_database.cc,backends/brass/brass_inverter.cc, backends/brass/brass_inverter.h: For allterms iteration of a WritableDatabase, if a prefix is specified only flush changes for terms with that prefix. Also, don't flush document lengths for allterms iteration. Tue Dec 22 22:48:55 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Fixed reversed checks for chert and flint being enabled in r13781. Tue Dec 22 14:20:24 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/testrunner.cc: Add brass variants of multi and remote backends. Tue Dec 22 14:15:43 GMT 2009 Olly Betts <olly@survex.com> * bin/: Implement xapian-check and xapian-compact support for brass. Tue Dec 22 13:41:08 GMT 2009 Olly Betts <olly@survex.com> * bin/: Split the chert-specific checking code into a separate file so xapian-check.cc is just the driver code. Tue Dec 22 12:47:59 GMT 2009 Olly Betts <olly@survex.com> * backends/brass/: Add new Inverter class to encapsulate the inversion data structures and associated code. Instead of carefully iterating modified posting lists and document length lists, just flush the particular list needed to disk and iterate that. This is much simpler and more obviously free of subtle bugs, and means we don't need to use data structures which allow such iteration. Tue Dec 22 12:25:04 GMT 2009 Olly Betts <olly@survex.com> * backends/brass/,backends/dbfactory.cc,include/xapian/dbfactory.h, include/xapian/valueiterator.h,include/xapian/version_h.cc, tests/harness/Makefile.mk: Fix to actually build brass. Testsuite doesn't currently all pass. Tue Dec 22 04:26:17 GMT 2009 Olly Betts <olly@survex.com> * HACKING,INSTALL,backends/Makefile.mk,backends/brass/, backends/chert/dir_contents,backends/databasereplicator.cc, backends/dbfactory.cc,common/pack.h,configure.ac,docs/,tests/, tests/harness/: Add new development backend called "brass" and promote "chert" to being the stable backend. Mon Dec 21 09:41:57 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc: Update comment about MAX_SAFE_TERM_LENGTH to reflect the key format changes. Mon Dec 21 08:12:55 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_alldocsmodifiedpostlist.cc, backends/chert/chert_alldocsmodifiedpostlist.h: Change reference parameter to const reference. Mon Dec 21 03:19:09 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/flint/flint_database.cc: Replacing a document deleted since the last flush failed to update the collection frequency and wdf, and caused an assertion failure when assertions were enabled. * tests/api_backend.cc: Add regression test replacedoc8. Sun Dec 20 07:26:30 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Divert to AS_MESSAGE_LOG_FD rather than literal fd 5. Divert more output to AS_MESSAGE_LOG_FD rather than /dev/null. Thu Dec 17 13:53:31 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Thu Dec 17 13:36:54 GMT 2009 Olly Betts <olly@survex.com> * Backport change from chert: * backends/flint/flint_alltermslist.cc: Tweak to not apply the prefix test in the "at end" case. * backends/flint/flint_alltermslist.cc: There's no point checking follow-on chunks match the prefix - the first chunk we'll hit which doesn't match has to be an initial chunk for a term. Thu Dec 17 11:57:31 GMT 2009 Olly Betts <olly@survex.com> * tests/api_replacedoc.cc: Fix to test the right variable. Thu Dec 17 11:16:51 GMT 2009 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: The regression test replacedoc5 no longer actually checks for the situation it was written for, due to the shortcutting we now do in flint and chert when replacing a document with itself, so extend it with a check with that behaviour suppressed (fortunately it still passes). Wed Dec 16 12:03:13 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/,bin/Makefile.mk,common/pack.h: Change the packing of uints and strings into sortable keys in chert databases, which reduces database size by 2.5% in tests. This means an incompatible change in the chert format. * bin/xapian-chert-update.cc: Utility to update a chert database from the old format to the new format. It works much like xapian-compact so should take a similar amount of time (and results in a compact database). * tests/api_wrdb.cc: Adjust the test which checks a string of zero bytes fails in the correct way - 126 zero bytes is now handled, so we need 127 to check the failure mode. Wed Dec 16 11:52:27 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Ensure that the resultant database has a fresh UUID (previously chert copied the UUID from the first input, while flint didn't set a UUID so one was generated on demand when next requested, if the database is writable. * tests/api_compact.cc: Add tests that a UUID is set and that it is different from that of the input. Wed Dec 16 09:57:27 GMT 2009 Olly Betts <olly@survex.com> * examples/quest.cc: If no database is specified, still parse the query and report get_description() on it as this provides a useful way to see how a query parses. Wed Dec 16 00:40:51 GMT 2009 Olly Betts <olly@survex.com> * languages/german.sbl,languages/german2.sbl: Update the german and german2 stemming algorithms to the latest versions from Snowball. These add an extra rule for the "-nisse" ending. Tue Dec 15 13:08:33 GMT 2009 Olly Betts <olly@survex.com> * bin/: Split apart the flint and chert compaction code. Should be no change in functionality. Tue Dec 15 10:11:10 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_alltermslist.cc: Tweak to not apply the prefix test in the "at end" case. Tue Dec 15 09:40:19 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_alltermslist.cc: There's no point checking follow-on chunks match the prefix - the first chunk we'll hit which doesn't match has to be an initial chunk for a term. Tue Dec 15 08:24:04 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/: Convert DerefStringWrapper_ to a templated DerefWrapper_ and use it instead of DocIDWrapper and TermPosWrapper. Tue Dec 15 08:23:08 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/query.h: Mark Query::Internal as @private for doxygen. Tue Dec 15 03:59:44 GMT 2009 Olly Betts <olly@survex.com> * common/pack.h: Return false not true if unpack_string_preserving_sort() fails. Add assertions that the source pointer isn't NULL to all the functions. Tue Dec 15 02:30:24 GMT 2009 Olly Betts <olly@survex.com> * examples/delve.cc: Improve phrasing ("unique terms" -> "distinct terms"). Sat Dec 12 13:07:51 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/Makefile.am,tests/api_posdb.cc,tests/api_replacedoc.cc: Split my recently added replacedoc test out into a new file. Sat Dec 12 12:58:22 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_posdb.cc: Extend test of document modification to cover removing the last term in the document, too. Sat Dec 12 12:57:43 GMT 2009 Richard Boulton <richard@tartarus.org> * backends/chert/chert_database.cc,backends/flint/flint_database.cc: Fix comparison of old term with new. Sat Dec 12 11:30:23 GMT 2009 Richard Boulton <richard@tartarus.org> * backends/chert/chert_database.cc,backends/flint/flint_database.cc, tests/api_posdb.cc: Delete old positionlist chunks in replace_document(). Uncomment part of test which tested this. Sat Dec 12 04:24:42 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/flint/flint_database.cc, common/document.h: If a document is replaced with itself unmodified, we no longer increase the automatic flush counter. * tests/api_backend.cc: Add regression test replacedoc7. Fri Dec 11 19:39:34 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_posdb.cc: Add a basic test of adding a document, and modifying its positions. Found one bug already, but the line which triggers this is commented out for now so that we don't break the snapshotters until it's fixed. Fri Dec 11 16:51:22 GMT 2009 Richard Boulton <richard@tartarus.org> * common/documentterm.h: Fix some documentation comments which were trivially incorrect or incomplete. Fri Dec 11 11:05:19 GMT 2009 Richard Boulton <richard@tartarus.org> * COPYING_MIT,HACKING: Add details of our policy for accepting patches. This has informally been true for a while, but it probably helps to tell people about it! Tue Dec 08 13:15:28 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.h: Fix to build with --enable-assertions. Tue Dec 08 13:05:36 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.cc,backends/chert/chert_cursor.h, backends/chert/chert_postlist.cc: Add MutableChertCursor subclass of ChertCursor and implement del() there rather than in ChertCursor. This means we can use the C++ type system to track whether we got a const or non-const ChertTable * and so whether it is safe to cast away that const and delete the entry the cursor points to. Tue Dec 08 12:03:08 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_table.cc, backends/chert/chert_table.h,backends/flint/flint_database.cc, backends/flint/flint_table.cc,backends/flint/flint_table.h, bin/xapian-compact.cc,bin/xapian-inspect.cc: Add empty() method to FlintTable and ChertTable which works completely reliably even if the item count has wrapped. * tests/api_closedb.cc: Add closedb3 for coverage of get_doccount() and has_positions(). Tue Dec 08 08:23:12 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.h: Tweak whitespace in comment. Mon Dec 07 04:56:42 GMT 2009 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_synonym.h: Eliminate "size" private member of FlintSynonymTermList which is set but never used. Mon Dec 07 04:49:54 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_synonym.h: Eliminate "size" private member of ChertSynonymTermList which is set but never used. Mon Dec 07 04:32:46 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/: Use > 32 bit type to keep count of items in a table. This fixes misreporting by xapian-check, and in the very unlikely case where the counter has just wrapped to 0, makes xapian-compact not mistakenly think tables are empty and the matcher ignore positional information. Thu Dec 03 10:08:28 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_compact.cc: Cast character value in string constructor to a char; without this we were getting a stack overflow and general nastiness on i386, which we hypothesise was due to the wrong overload for string() being picked. Wed Dec 02 23:10:50 GMT 2009 Olly Betts <olly@survex.com> * backends/flint_lock.cc: Add missing include of xapian/error.h. This was causing the snapshot builder to fail, but worked OK in my tree for reasons I don't understand. Wed Dec 02 14:30:40 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc, backends/chert/chert_databasereplicator.cc, backends/flint/flint_database.cc, backends/flint/flint_databasereplicator.cc,backends/flint_lock.cc, backends/flint_lock.h: Factor out 4 copies of the same code to report why Xapian failed to get a lock. Wed Dec 02 10:05:27 GMT 2009 Olly Betts <olly@survex.com> * backends/Makefile.mk,backends/chert/,backends/flint/, backends/flint_lock.cc,backends/flint_lock.h: Factor out a single version of the "flint-compatible locking code" and use it from both flint and chert. Wed Dec 02 02:49:14 GMT 2009 Olly Betts <olly@survex.com> * tests/api_compact.cc,tests/apitest.cc,tests/apitest.h, tests/harness/backendmanager.cc,tests/harness/backendmanager.h: Add get_database_path() variant which takes a generator function and use it so that the right database names are used in compactnorenumber1. Tue Dec 01 12:37:24 GMT 2009 Olly Betts <olly@survex.com> * tests/apitest.cc,tests/harness/backendmanager.cc, tests/harness/backendmanager.h: Move generator function variant of get_database() to BackendManager, and tweak it to build the database under a temporary path and rename it once built, to avoid leaving a partial database in place if interrupted. * tests/perftest/perftest_matchdecider.cc: Use this mechanism to lazily generate test databases. * tests/harness/: Remove the optional name parameter from various BackendManager methods as it was only used by perftest and is no longer needed. Tue Dec 01 11:00:35 GMT 2009 Olly Betts <olly@survex.com> * tests/apitest.cc,tests/apitest.h: Add mechanism for caching databases generated by a function. * tests/api_compact.cc: Use it for compactnorenumber1. Tue Dec 01 10:59:50 GMT 2009 Olly Betts <olly@survex.com> * tests/api_db.cc: Rename matchfunctor<n> to matchdecider<n> to match current terminology. Tue Dec 01 10:51:23 GMT 2009 Olly Betts <olly@survex.com> * tests/api_db.cc: Don't run matchfunctor3 under remote backends as MatchDecider isn't actually supported there (uncovered by recent change to throw UnimplementedError in this case). Tue Dec 01 09:35:53 GMT 2009 Richard Boulton <richard@tartarus.org> * include/xapian/matchspy.h: Remove out-of-date reference to add_slot(), which no longer exists: users are expected to use multiple ValueCountMatchSpies if they need to monitor more than one slot. Tue Dec 01 08:04:17 GMT 2009 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/api_compact.cc: Add test coverage for xapian-compact --no-renumber with multiple databases. Tue Dec 01 06:25:47 GMT 2009 Olly Betts <olly@survex.com> * tests/api_backend.cc: Clean up matchdecider4. Tue Dec 01 06:12:29 GMT 2009 Olly Betts <olly@survex.com> * HACKING: A couple more std::string efficiency tips. Tue Dec 01 05:51:52 GMT 2009 Olly Betts <olly@survex.com> * api/omenquire.cc,common/multimatch.h,matcher/multimatch.cc, net/remoteserver.cc: Throw UnimplementedError if a matchdecider or "legacy" matchspy is used with the remote backend. * tests/api_backend.cc: Add regression test matchdecider4. Mon Nov 30 05:42:43 GMT 2009 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remote-database.h, common/remoteprotocol.h,common/remoteserver.h, docs/remote_protocol.html,net/remoteserver.cc,tests/api_metadata.cc, tests/harness/testrunner.cc: Add support for WritableDatabase::set_metadata() and Database::get_metadata() to the remote backend (based largely on patch in #178). * AUTHORS: Thanks to Paul Rudin for patch. Thu Nov 26 07:25:15 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Start to update from ChangeLog. Thu Nov 26 02:26:43 GMT 2009 Olly Betts <olly@survex.com> * common/str.cc: Tweak the length calculation for the buffer used by str() when converting integers to strings so that the buffer is exactly the minimum required size for integers of size 1, 2, 4, and 8 bytes, and only a byte extra for 16 byte integers. Wed Nov 25 05:59:13 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: In the --no-renumber case, just set offset[] entries to zero to start with rather than fixing them up later. Wed Nov 25 05:49:37 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Extend --no-renumber to support merging databases, but only if they have disjoint ranges of used document ids. Wed Nov 25 04:37:38 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Fix typos in --help output. Wed Nov 25 04:22:26 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Prune unused docids off the end of database when merging multiple databases with renumbering. Wed Nov 25 03:12:24 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Use string() instead of "" and string(1, '\0') instead of string("", 1). Wed Nov 25 02:34:10 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_lock.cc: Fix locking code to work if stdin and/or stdout have been closed. * backends/flint/flint_lock.cc: Backport fix to flint. * tests/api_backend.cc: Add regression test lockfilefd0or1. Wed Nov 25 00:59:05 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_lock.cc, backends/chert/chert_lock.h: Add "FDLIMIT" reason code for why locking failing. Tue Nov 24 12:44:56 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc: Factor out throwing an exception for docid 0 not being valid. Tue Nov 24 12:39:10 GMT 2009 Olly Betts <olly@survex.com> * tests/api_none.cc: Note previous fix was ticket#415. Tue Nov 24 12:33:12 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc: Fix many Xapian::Database methods to behave better on a database with no subdatabases, such as is constructed by Database(). * tests/Makefile.am,tests/api_nodb.cc,tests/api_none.cc: Expand emptydb_metadata1 to also regression test these other cases, and rename to nosubdatabases1. Mon Nov 23 14:01:51 GMT 2009 Olly Betts <olly@survex.com> * common/debuglog.h: Add "NO_ARGS" which can be used for debug logging when a function or method takes no arguments. The main advantage over "" is that no extra code is generated for it. * api/omdatabase.cc: Convert to use debuglog.h fully. * api/matchspy.cc,api/omdatabase.cc: Make use of NO_ARGS. Mon Nov 23 09:10:55 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.cc,backends/flint/flint_cursor.cc, common/debuglog.cc,common/debuglog.h,common/omdebug.h,configure.ac, matcher/multimatch.cc,queryparser/queryparser.lt: Rename XAPIAN_DEBUG_VERBOSE macro to XAPIAN_DEBUG_LOG to better match the user-visible "--enable-log" configure option which controls it. Mon Nov 23 08:08:45 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_lock.cc, backends/chert/chert_table.cc,backends/flint/: Use string() or s.resize(0) instead of "". Use string(1, '\0') instead of string("", 1). Sun Nov 22 14:20:51 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_postlist.cc: When skipping through a chunk of postings to find the one we want, don't bother to unpack the wdf values we're skipping over, which should save a significant amount of time in certain cases where the profile data shows we spend something like 1/3 of the time in the function where this happens. Sun Nov 22 13:12:39 GMT 2009 Olly Betts <olly@survex.com> * api/matchspy.cc,backends/chert/chert_values.cc, languages/steminternal.cc: Tweak whitespace after while to be consistent. Sat Nov 21 17:39:23 GMT 2009 Richard Boulton <richard@tartarus.org> * backends/chert/chert_values.cc,backends/chert/chert_values.h: Make ValueChunkReader::skip_to() assign the value to a string only when the target has been reached. Saves a lot of unnecessary string copying - on a benchmark of mydeco data, improves time for 100 queries from 3.66s to 3.10s. Sat Nov 21 01:13:26 GMT 2009 Olly Betts <olly@survex.com> * tests/: Make arrays which don't need to be modified const. Sat Nov 21 00:01:04 GMT 2009 Olly Betts <olly@survex.com> * tests/api_matchspy.cc: The default parameter for get_writable_database() is string(), so don't explicitly pass "". Fri Nov 20 16:19:13 GMT 2009 Richard Boulton <richard@tartarus.org> * tests/api_matchspy.cc: Add regression test for underflow in numeric ranges (based on python version in ticket #321). Fri Nov 20 10:54:08 GMT 2009 Olly Betts <olly@survex.com> * matcher/valuestreamdocument.cc: Use check() instead of skip_to() in ValueStreamDocument. Fri Nov 20 09:57:12 GMT 2009 Olly Betts <olly@survex.com> * common/document.h,matcher/: Implement ValueStreamDocument subclass of Xapian::Document::Internal which provides a way to connect up valuestreams to uses of values during the match. Other (non-recommended) uses of the Document passed to MatchDecider, etc are passed through to a lazily created backend Document::Internal subclass. This should be a lot more efficient for chert. It's unclear how it will affect performance for backends which don't store values in streams - profiling is required. Fri Nov 20 02:35:14 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_document.cc: Fix closedb1 on InMemory backend, broken by recent commit. Thu Nov 19 12:03:52 GMT 2009 Olly Betts <olly@survex.com> * backends/multi/multi_valuelist.cc: Fix to handle the case where all the sublists turn out to be empty when we first call next(). Thu Nov 19 11:12:53 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/: Set the document data and values lazily for the inmemory backend too. They're much less costly to fetch than if a disk access may be needed, but it avoids a copy of the respective data if they aren't needed, and either could potentially be large. Consistency here also makes things easier to understand. Thu Nov 19 08:47:24 GMT 2009 Olly Betts <olly@survex.com> * api/omdocument.cc: Prefer string() to "". Thu Nov 19 07:38:29 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/multi/Makefile.mk, backends/multi/multi_valuelist.cc,common/Makefile.mk, common/multivaluelist.h,tests/api_valuestream.cc: Implement support for iterating valuestreams for multidatabases. Thu Nov 19 06:19:05 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Rework the checking of postlist chunks into a cleaner approach which should report errors better, and eliminate the (probably) impossible "did might not be initialised" path which GCC 4.2 spotted. Thu Nov 19 01:15:03 GMT 2009 Olly Betts <olly@survex.com> * common/pack.h: Remove incorrect assertion and document what NULL parameter value means in this case. Wed Nov 18 16:33:01 GMT 2009 Richard Boulton <richard@tartarus.org> * bin/xapian-check.cc: Fix compiler warning (with gcc-4.2) that did could be used uninitialised: it couldn't be, but it's hard for a compiler to see that. Wed Nov 18 12:48:17 GMT 2009 Richard Boulton <richard@tartarus.org> * AUTHORS: Update my email address. Wed Nov 18 12:21:09 GMT 2009 Olly Betts <olly@survex.com> * common/valuelist.h: Fix internal documentation comment. Wed Nov 18 10:40:54 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Use <cstdlib>, <cmath>, <cstdio>, <cstring> for configure tests since we use these forms in the code and want consistency between what configure tests and what the code uses. * NEWS,configure.ac: Update for 1.1.3. Wed Nov 18 10:39:59 GMT 2009 Olly Betts <olly@survex.com> * PLATFORMS: Update from 1.0.17. Wed Nov 18 10:31:49 GMT 2009 Olly Betts <olly@survex.com> * INSTALL: Improve text about zlib dependency. Wed Nov 18 10:13:59 GMT 2009 Olly Betts <olly@survex.com> * INSTALL: Note the package to install for building against libuuid on Fedora, and note that libuuid isn't required on Microsoft Windows. Wed Nov 18 02:34:30 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from 1.0.17 and ChangeLog. Mon Nov 16 08:35:52 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix interaction of FLAG_PARTIAL and FLAG_SYNONYM. (ticket#407) * tests/Makefile.am,tests/api_qpbackend.cc: Add regression test qpsynonympartial1. Mon Nov 16 02:15:14 GMT 2009 Olly Betts <olly@survex.com> * backends/flint/flint_version.cc: Add "using namespace std;" and drop the explicit std:: qualifiers. Fri Nov 13 02:27:48 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_spelling.h,backends/chert/chert_table.h, backends/flint/flint_spelling.h,backends/flint/flint_table.h, common/getopt.cc,common/stringutils.h,languages/steminternal.h: Add missing std:: qualifiers for functions in <cstring>, <cstdlib>, etc. Fri Nov 13 02:25:50 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_lock.cc,backends/flint/flint_lock.cc: Add "using namespace std;" and drop the explicit std:: qualifiers. Fri Nov 13 02:11:18 GMT 2009 Olly Betts <olly@survex.com> * common/str.cc: Use assignment to initialise a size_t - the current code was turning a missing snprintf() prototype into a confusing error message with Sun's C++, and is less clear anyway. Fri Nov 13 02:03:28 GMT 2009 Olly Betts <olly@survex.com> * HACKING,backends/chert/chert_lock.cc,backends/flint/flint_lock.cc, backends/remote/remote-database.cc,net/remoteserver.cc, net/tcpserver.cc,tests/harness/backendmanager_remotetcp.cc, tests/harness/testsuite.cc: Sun C++'s <csignal> (not unreasonably) excludes non-ISO-C functions which are in <signal.h> - for example, kill() (which POSIX specifies). So revert <csignal> to <signal.h> and document this as a global exception. Thu Nov 12 08:30:07 GMT 2009 Olly Betts <olly@survex.com> * matcher/collapser.h: Remove spurious comma after last entry in enum, which Sun's C++ compiler warns about. Thu Nov 12 00:24:55 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Sorter -> KeyMaker in an exception message. Thu Nov 12 00:01:55 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/database.h: Tweak documentation comments about committing changes. Wed Nov 11 12:23:48 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix handling of a group of stopwords which notably caused issues when default_op was OP_AND, but could probably manifest in other cases too. Fixes ticket#406. * tests/queryparsertest.cc: Add regression test qp_stopword_group1. Wed Nov 11 10:54:58 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_version.h,common/safeuuid.h: Fix workaround for uuid.h headers which lack const to actually work on Solaris. Wed Nov 11 10:53:50 GMT 2009 Olly Betts <olly@survex.com> * common/Makefile.mk: Need to ship new pack.h header. Tue Nov 10 22:27:47 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Fix compiler warning introduced by previous change. Tue Nov 10 12:24:34 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/,bin/xapian-check.cc,common/pack.h: Rewrite the packing and unpacking functions more efficiently. As well as being generally faster, the pack functions now take a reference to a string to append to, which avoids creating a lot of temporary string objects. Indexing HTML files with omindex is 5-10% faster. Searching for "The" on gmane (which results in a lot of unpacking of postings and document lengths) is about 35% faster. (ticket#326) Thu Nov 05 03:55:37 GMT 2009 Olly Betts <olly@survex.com> * common/safeuuid.h: Fix to work with uuid libraries which lack const qualifiers. Fixes build failure on Solaris. Thu Nov 05 00:15:24 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Update details of debian packaging. Wed Nov 04 04:11:38 GMT 2009 Olly Betts <olly@survex.com> * common/output.h: Use C++ syntax for NULL with a type in log output. Wed Nov 04 04:02:09 GMT 2009 Olly Betts <olly@survex.com> * queryparser/lemon.c: Avoid trivial memory leak. Wed Nov 04 00:13:00 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/database.h: Reword to avoid somewhat ambiguous "this method". Tue Nov 03 12:12:09 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/database.h: Make it more clear that you need to add the unique term to the document if you want to use replace_document() to use external unique IDs for documents. Tue Nov 03 06:08:43 GMT 2009 Olly Betts <olly@survex.com> * tests/generate-api_generated: Update MultiValueSorter to MultiValueKeyMaker. Tue Nov 03 05:51:23 GMT 2009 Olly Betts <olly@survex.com> * tests/api_sorting.cc,tests/api_sortingold.cc: Fix changesorter1 and oldchangesorter1 to not be run for the remote backend where they will now fail with UnimplementedError. Tue Nov 03 02:43:12 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Xapian::Sorter isn't supported with the remote backend so throw UnimplementedError rather than giving incorrect results. (ticket#384) * tests/api_sorting.cc: Add sortfunctorremote1 test this exception is actually thrown. Tue Nov 03 01:26:57 GMT 2009 Olly Betts <olly@survex.com> * common/database.h: Fix vertical whitespace glitch. Mon Nov 02 08:10:03 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc: Check that internal is not empty before referencing its first element, in get_metadata() and metadata_keys_begin(), to avoid a potential segfault. * tests/api_nodb.cc: Add emptydb_metadata1, a regression test for this bug. Sun Nov 01 22:10:54 GMT 2009 Olly Betts <olly@survex.com> * examples/delve.cc,examples/quest.cc: Extend exception handling to the whole of main. Xapian::Stem("english") can't actually throw, but that's not obvious to static analysis tools, and it is more robust to wrap the whole of main, and reduces indentation. Sun Nov 01 21:27:05 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Add missing exception catch for const char *. The only case which currently throws this is an "impossible" situation, but if we're going to check for it, the reporting of failure should actually work. Identified by Coverity's Scan. Sun Nov 01 07:56:38 GMT 2009 Olly Betts <olly@survex.com> * examples/quest.cc: Tighten up the type of the error we catch to detect an unknown stemming language. Sat Oct 31 07:16:11 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Sat Oct 31 06:59:19 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Fix deprecation warnings when building with recent GCC. Sat Oct 31 02:46:04 GMT 2009 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add another test query string to cover a case we didn't previously check. Tue Oct 27 00:42:59 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_lock.cc,backends/flint/flint_lock.cc: Don't try to close the fd one more than the maximum allowable. (ticket#408) * AUTHORS: Add Carl Worth for the patch. Mon Oct 05 09:00:58 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Mention e2fsprogs-devel in the message thrown up by configure, too. Mon Oct 05 08:53:05 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * HACKING: Note the command needed to get the uuid library installed on fedora, since it's rather non-obvious. Sat Oct 03 20:11:48 GMT 2009 Olly Betts <olly@survex.com> * unicode/tclUniData.cc: Update Unicode character database to Unicode 5.2. (ticket#351) * tests/api_unicode.cc: Add tests for some characters added in Unicode 5.2. Thu Oct 01 21:19:34 GMT 2009 Olly Betts <olly@survex.com> * xapian-config.in: Need to quote ^ for Solaris /bin/sh. Fri Sep 18 13:23:00 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Actually use any flags we determine are needed to switch the compiler to proper ANSI C++ mode, when building xapian-core - this stopped working in 1.0.12, breaking support for HP's aCC, Compaq's cxx, Sun's CC, and SGI's CC. Fri Sep 18 12:40:37 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/matchspy.h: Mark ValueCountMatchSpy, NumericRange, NumericRanges, and score_evenness() as experimental. Fri Sep 18 08:36:34 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/weight.h,weight/weight.cc: Remove default implementation of Weight::clone() which returns NULL - we always need clone() to be implemented because it's called for every term in the query, not just used for the remote backend. Fri Sep 18 08:02:08 GMT 2009 Olly Betts <olly@survex.com> * api/omenquire.cc,docs/deprecation.rst,include/xapian/enquire.h: Mark and document the matchspy parameter of Enquire::get_mset() as deprecated in favour of the new MatchSpy class and Enquire::add_matchspy(). Fri Sep 18 07:58:11 GMT 2009 Olly Betts <olly@survex.com> * tests/perftest/perftest_matchdecider.cc: Don't pass default values for optional parameters to Enquire::get_mset(). Fri Sep 18 02:36:43 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: If FLAG_PARTIAL is specified, don't try to spell correct a term at the end of the query which we attempt to expand as partial. * tests/queryparsertest.cc: Add regression test qp_spellpartial1. Fri Sep 18 02:33:55 GMT 2009 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: For testcase qp_spellwild1: Fix the dbdir to qp_spellwild1 to match the testname; Fix the ordering of TESTCASE macros to match the order of the test functions; Note that this is a regression test and the versions the bug will be fixed in. Fri Sep 18 01:25:37 GMT 2009 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Fix to work with the old scoping rules for variables declared in for(). Thu Sep 17 19:17:30 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.lemony: Don't apply spelling correction to wildcard terms if FLAG_WILDCARD and FLAG_SPELLING_CORRECTION are both specified. * tests/queryparsertest.cc: Add qp_spellwild1 as a test for this. Thu Sep 17 12:08:17 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Thu Sep 17 10:20:35 GMT 2009 Olly Betts <olly@survex.com> * api/keymaker.cc: Implement Richard's idea of not encoding any trailing, forward sorted, empty values which is a generalisation of encoding all forward sorted, empty values as an empty string. * tests/api_sorting.cc: Add new testcase multivaluekeymaker1 to check encodings are as expected. Thu Sep 17 07:15:10 GMT 2009 Olly Betts <olly@survex.com> * api/,common/multimatch.h,common/omenquireinternal.h, docs/deprecation.rst,docs/sorting.rst,include/Makefile.mk, include/xapian.h,include/xapian/enquire.h,include/xapian/keymaker.h, include/xapian/sorter.h,matcher/multimatch.cc,tests/Makefile.am, tests/api_sorting.cc,tests/api_sortingold.cc: Rename Sorter to KeyMaker, paving the way for using it to build collapse keys too. Resolve the inconsistency in MultiValueSorter::add()'s "forward" parameter by replacing it with MultiKeyMaker::add_value() with a "reverse" parameter. (ticket#359) Thu Sep 17 07:13:25 GMT 2009 Olly Betts <olly@survex.com> * api/valuesetmatchdecider.cc,include/xapian/valuesetmatchdecider.h: Inline trivial constructor from header. Add explicit '#include "xapian/document.h"'. Thu Sep 17 05:22:50 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Document how to use the XAPIAN_DEPRECATED() macro to mark a class as deprecated. Wed Sep 16 02:40:13 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: Eliminate a couple of assertions which can never fire. Sat Sep 12 04:05:35 GMT 2009 Olly Betts <olly@survex.com> * matcher/msetpostlist.cc: Fix MSetPostList not to read off the end of the MSet if get_maxweight() is called when at_end(). This can happen but the testcase was too large to send so sadly no regression test I'm afraid. Sat Sep 12 03:41:04 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_values.cc: Fix code to find the first docid in the net chunk (ticket#399). * tests/api_backend.cc: Add regression test from Rich Lane. Fri Sep 11 11:14:49 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/,bin/xapian-check.cc,bin/xapian-compact.cc: Add support for chert databases without a termlist table (ticket#181). Currently the only way to create such a database is to create a chert database and do "rm termlist.*". Fri Sep 11 10:07:35 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Don't report an absent lazy input table as 0 size. Fri Sep 11 03:17:17 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_table.h: Improve documentation of "handle" member. Thu Sep 10 13:24:29 GMT 2009 Olly Betts <olly@survex.com> * m4-macros/xapian-1.1.m4: Report the default xapian-config basename in configure's --help output. Thu Sep 10 12:55:48 GMT 2009 Olly Betts <olly@survex.com> * m4-macros/xapian-1.1.m4: Add optional third parameter to XO_LIB_XAPIAN which specifies the basename for the "xapian-config" script (defaults to "xapian-config" to give the current behaviour). Thu Sep 10 06:39:17 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update with changes in 1.0.16 and from ChangeLog. Thu Sep 10 02:17:56 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Clarify the documentation comments for QueryParser::set_default_op() and QueryParser::get_default_op(). Wed Sep 09 13:36:58 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/andmaybepostlist.cc,tests/Makefile.am,tests/api_queryopt.cc: Back out patch from r13440 which introduced a new optimisation for AND_MAYBE when the maximum weight of the RHS becomes zero, due to various undesirable side effects. I've put the patch into ticket #400 with an explanation. Wed Sep 09 01:54:05 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_postlist.cc,tests/api_backend.cc: Fix WritableDatabase::get_doclength() to work properly after a call to commit for the chert backend (ticket#397). Wed Sep 09 00:28:09 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS: Add Dmitry Liakh for fix for #398. Wed Sep 09 00:02:47 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_lock.cc,backends/flint/flint_lock.cc: Fix a typo which stopped this fix in 1.0.12 from working (ticket #398): If we fail to get the lock after we spawn the child lock process (the common case is because the database is already open for writing) then we now clean up the child process properly. Fri Sep 04 11:24:26 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * docs/categorisation.rst: Remove out-of-date call to get_mset() which passed a MatchSpy in - spies are now set with add_matchspy() before get_mset() is called. Thu Sep 03 00:29:27 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/andmaybepostlist.cc: For AndMaybe, if the RHS has a maximum possible weight of 0, we now decay to just the LHS. * tests/Makefile.am,tests/api_queryopt.cc: Test coverage that this optimisation doesn't break anything. Tue Sep 01 12:04:13 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Send stderr output from ldconfig to config.log. Tue Sep 01 10:35:51 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS,common/safeuuid.h: Fix uuid_unparse_lower() replacement for older libuuid to actually compile (really fixes ticket#368). Mon Aug 31 05:58:55 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Sync with 1.0.15. Mon Aug 31 05:48:13 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Start to format entries for 1.1.3. Mon Aug 31 04:14:58 GMT 2009 Olly Betts <olly@survex.com> * PLATFORMS: Sync with 1.0.15. Mon Aug 31 01:30:01 GMT 2009 Olly Betts <olly@survex.com> * common/getopt.cc: Use USE_GLIBC_GNUGETOPT from gnu_getopt.h rather than repeating the conditionals used to determine it here. Rename __getopt_initialized to getopt_initialized and make it static. Rename _getopt_initialize() to getopt_initialize(). Mon Aug 31 01:24:29 GMT 2009 Olly Betts <olly@survex.com> * common/gnu_getopt.h: Update (C) dates. Markup file description for doxygen. Fix comment reference to ctype.h as we now use <cctype>. Sun Aug 30 23:28:25 GMT 2009 Olly Betts <olly@survex.com> * common/gnu_getopt.h: Make optarg, optind, opterr, and optopt extern "C" to avoid linkage clash with these symbols on Mac OS X 10.6. Thu Aug 27 03:05:33 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Add note to document the reason for any exceptions to the rule to use C++ forms of ISO C headers. Wed Aug 26 13:07:07 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS,INSTALL,docs/install.html: Drop .php from xapian.org URLs. Wed Aug 26 12:58:51 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Update URLs. Remove duplicated text about updating RoadMap from the release checklist. Wed Aug 26 10:56:07 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Update various URLs. Tue Aug 25 01:52:51 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Fix not to report heaps of bogus errors under valgrind 3.5.0. Mon Aug 24 02:19:06 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Also update the documentation comment for set_sort_by_relevance_then_value() to mention sortable_serialise(). Fri Aug 21 14:35:35 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * Makefile.am: Add generate-exceptions to EXTRA_DIST - was missing from tarballs. Fri Aug 21 12:21:59 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * AUTHORS: Add Michael Vogt, who suggested adding a comment about sortable_serialise to set_sort_by_value(). Fri Aug 21 12:14:13 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/enquire.h: Update documentation comments for set_sort_by_value() and set_sort_by_value_then_relevance() to mention sortable_serialise() as a good way to store values if you want to sort by them. Mon Aug 17 16:51:10 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: When sorting by non-pure-relevance, ensure that the document is shown to the matchspy even if it couldn't get in the mset. * tests/api_matchspy.cc: Extend matchspy4 to include a regression test for this. Mon Aug 17 11:51:24 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/matchspy.cc,include/xapian/matchspy.h: Add overload for score_evenness to allow a NumericRanges object to be passed to it. Mon Aug 17 07:36:57 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/matchspy.cc,include/xapian/matchspy.h: Replace build_numeric_ranges() function with a NumericRanges object, whose constructor is passed the values and which exposes the calculated ranges via a getter. * tests/api_matchspy.cc: Update tests to use new NumericRanges object instead of build_numeric_ranges(). Wed Aug 12 17:33:53 GMT 2009 Olly Betts <olly@survex.com> * Makefile.am: Fix generate-exceptions change to work in a VPATH build. Wed Aug 12 15:57:44 GMT 2009 Olly Betts <olly@survex.com> * exception_data.pm: Add new SerialisationError. * common/serialise-double.cc: Throw SerialisationError or InternalError instead of NetworkError. Wed Aug 12 15:19:05 GMT 2009 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,generate-exceptions,generate-exceptions.in: It's not very useful to be able to run generate-exceptions outside the build system, and it only gets the #! line substituted, so just move those runes into Makefile.am and reduce the number of files that configure needs to generate by one. Mon Aug 10 13:25:22 GMT 2009 Olly Betts <olly@survex.com> * api/matchspy.cc: Use <cfloat> and <cmath> rather than <float.h> and <math.h> (regression from merging the matchspy branch). Mon Aug 10 13:01:57 GMT 2009 Olly Betts <olly@survex.com> * api/,bin/xapian-tcpsrv.cc,common/,docs/postingsource.rst, docs/serialisation.rst,include/Makefile.mk,include/xapian.h, include/xapian/,net/remoteserver.cc,tests/: Rename SerialisationContext to Registry. Mon Aug 10 11:23:23 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Start to update from ChangeLog. Mon Aug 10 10:29:13 GMT 2009 Olly Betts <olly@survex.com> * api/omenquire.cc,include/xapian/weight.h,matcher/localmatch.cc: Make Weight::clone() public, and remove Weight::clone_() as it is no longer required. Mon Aug 10 07:55:04 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * HACKING: Add section on ordering of include files, and fix a few formatting issues to make this file into valid restructured text again. Mon Aug 10 06:55:45 GMT 2009 Olly Betts <olly@survex.com> * api/serialisationcontext.cc,common/serialisationcontextinternal.h: Refactor to use a template function to implement the exception safe approach I implemented for PostingSource, and use it for MatchSpy and Weight too. * tests/api_serialise.cc: Add tests for exception safety of MatchSpy and Weight with SerialisationContext. Mon Aug 10 06:15:23 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/weight.h,weight/weight.cc: Add default implementations of Weight methods name(), serialise(), unserialise(), and clone() for consistency with PostingSource and MatchSpy. Mon Aug 10 03:29:12 GMT 2009 Olly Betts <olly@survex.com> * tests/api_postingsource.cc: Tweak header include order. Tue Aug 04 15:49:54 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Explicitly document that an empty prefix argument to QueryParser::add_prefix() means "no prefix". Tue Aug 04 13:11:26 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc: Throw UnimplementedError rather than InvalidOperationError from PostingSource::serialise() and PostingSource::unserialise() for consistency with MatchSpy. * include/xapian/postingsource.h: Document this behaviour (previously we didn't say what the default implementation actually did). * tests/api_serialise.cc: Add test coverage. Mon Aug 03 11:40:31 GMT 2009 Olly Betts <olly@survex.com> * api/serialisationcontext.cc: Fix to handle the clone() method or dtor of a PostingSource subclass throwing exceptions. * tests/api_serialise.cc: Add new testcase serialisationcontext1 as a regression test for this. Mon Aug 03 06:40:28 GMT 2009 Olly Betts <olly@survex.com> * tests/api_serialise.cc: Fix copy-and-pasted @brief. Sun Aug 02 16:21:54 GMT 2009 Olly Betts <olly@survex.com> * common/remoteprotocol.h: Correction: protocol version 33 will debut in 1.1.3 not 1.1.4. Sun Aug 02 16:20:28 GMT 2009 Olly Betts <olly@survex.com> * docs/remote_protocol.html: Update to document protocol version 33. Sun Aug 02 16:17:24 GMT 2009 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,net/remoteserver.cc: We were ignoring any trailing junk after the matchspies in MSG_GETMSET, so change to not send a count of them and just unpack until we run out of data. Sun Aug 02 15:47:30 GMT 2009 Olly Betts <olly@survex.com> * common/remoteprotocol.h: Note the versions in which recent protocol versions were introduced. Sun Aug 02 15:09:13 GMT 2009 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/serialise.h, net/remoteserver.cc,net/serialise.cc: Put the serialised MSet last in REPLY_RESULTS as then it doesn't need to know its own length (so the encoding is the same size it used to be in the non-matchspy case), and we avoid having to check for there being "junk" data left over after unserialising. Sun Aug 02 14:39:18 GMT 2009 Olly Betts <olly@survex.com> * matcher/remotesubmatch.h: Fix to build with GCC 4.4.1. Sun Aug 02 14:27:44 GMT 2009 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc: Don't assign a temporary string object. Sun Aug 02 13:50:14 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/matchspy.h: Add missing documentation comments. Sun Aug 02 12:15:43 GMT 2009 Richard Boulton <richard@lemurconsulting.com> Merge from matchspy branch. * api/,backends/remote/remote-database.cc,common/,docs/Makefile.am, docs/categorisation.rst,include/Makefile.mk,include/xapian.h, include/xapian/enquire.h,include/xapian/matchspy.h, include/xapian/serialisationcontext.h,matcher/multimatch.cc, matcher/remotesubmatch.cc,matcher/remotesubmatch.h, net/remoteserver.cc,net/serialise.cc,tests/Makefile.am, tests/api_matchspy.cc: Add new-style Xapian::MatchSpy class, which is a pure "spy" class, rather than being able to take a decision on whether a spy is applied before or after a result. This class is also designed to work with remote databases, passing the results back in serialised form. Also, add ValueCountMatchSpy, which counts the occurrences of each value in a slot in the search results seen (useful for faceted or categorisation systems). Fri Jul 31 12:50:57 GMT 2009 Olly Betts <olly@survex.com> * tests/api_backend.cc,tests/api_wrdb.cc: Move new test modifiedpostlist1 from api_wrdb.cc to api_backend.cc - we're trying not to make api-wrdb.cc any larger. Also note the ticket number and the release this will be fixed in. Thu Jul 30 16:30:06 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_modifiedpostlist.cc: Fix ChertModifiedPostList to skip added-but-then-deleted-before-flush documents. (ticket#392) backends/flint/flint_modifiedpostlist.cc: Same fix for flint. * tests/api_wrdb.cc: Add regression test modifiedpostlist1. * AUTHORS: Add Rich Lane for patch. Wed Jul 29 09:44:15 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Wed Jul 29 09:29:45 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/valueiterator.h: Fix typos ("again" -> "and"). Wed Jul 29 06:18:41 GMT 2009 Olly Betts <olly@survex.com> * tests/Makefile.am: Actually distribute testdata/apitest_declen.txt - new test data from the postingsources branch merge. Tue Jul 28 16:38:55 GMT 2009 Richard Boulton <richard@lemurconsulting.com> Merge from postingsources branch: * tests/generate-api_generated: Test get_description() methods of Xapian::PostingSource subclasses. Tue Jul 28 16:35:54 GMT 2009 Richard Boulton <richard@lemurconsulting.com> Merge from postingsources branch: * api/Makefile.mk,api/decvalwtsource.cc,api/serialisationcontext.cc, include/xapian/postingsource.h: Add DecreasingValueWeightPostingSource class, which reads weights from a value slot in which a significant range of the values are in decreasing order. This functions similarly to ValueWeightPostingSource, but can be much more efficient. * tests/api_valuestream.cc,tests/testdata/apitest_declen.txt: Tests, and some associated constructed test data, for DecreasingValueWeightPostingSource. Mon Jul 27 04:50:45 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Now using autoconf 2.64. Mon Jul 27 04:46:17 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Add note about preferring std::string::assign(), and also one about building up strings using +=. Mon Jul 27 04:45:15 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_lock.cc,backends/chert/chert_utils.h, backends/flint/flint_lock.cc,backends/flint/flint_utils.h, backends/remote/remote-database.cc: Use std::string::assign() rather than constructing a temporary string object to assign. Sun Jul 26 16:08:10 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS: Update for bugs fixed in 1.0.14. Sun Jul 26 16:03:54 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Update the release checklist. Sun Jul 26 14:28:25 GMT 2009 Olly Betts <olly@survex.com> * docs/doxygen_source.conf.in: Update roughly in-line with apidoc changes. I've not looked at the output much though. Sun Jul 26 14:00:14 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/database.h: Remove documentation comment for namespace Xapian - this is now documented in xapian.h. Sun Jul 26 13:59:42 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Add documentation comment for namespace Unicode. Sun Jul 26 13:58:44 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/dbfactory.h: Add documentation comments for the backend namespaces (InMemory, etc). Sun Jul 26 13:57:29 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: Add documentation comments for the generated version.h, and all the preprocessor defines it contains. Sun Jul 26 13:56:15 GMT 2009 Olly Betts <olly@survex.com> * include/xapian.h: Consistently say "linked with" rather than "linked to" in some cases. Put documentation comment for namespace Xapian in this file. Sun Jul 26 13:06:45 GMT 2009 Olly Betts <olly@survex.com> * docs/doxygen_api.conf.in: More fettling to try to get better output: + Wrap long comments. + Quote arguments containing @...@ substitutions as they could conceivably contain spaces. + ALWAYS_DETAILED_SEC = NO: since a detailed section with no extra information is pointless. + STRIP_FROM_PATH, INCLUDE_PATH: Add "@top_builddir@/include" for VPATH builds. + HIDE_UNDOC_MEMBERS = YES: otherwise @internal members show up (with no documentation). + HIDE_FRIEND_COMPOUNDS = YES: as friend declarations in the API headers are implementation details. + SHOW_INCLUDE_FILES = NO: users should just include <xapian.h> for all their API needs. + GENERATE_TODOLIST, GENERATE_TESTLIST, GENERATE_BUGLIST: all set to NO as we don't want these in the API documentation. + SHOW_NAMESPACES = YES: We have several sub-namespaces so these are useful to document. + INPUT: Prepend "@top_builddir@/include/xapian/version.h" so that XAPIAN_HAS_FLINT_BACKEND, etc are defined before parsing other headers. + EXCLUDE: remove all entries and use EXCLUDE_PATTERNS instead. + EXCLUDE_PATTERNS: Include entries corresponding to those in EXCLUDE and add */derefwrapper.h and */.* (the latter avoids trying to look in .svn or any other hidden files or directories). + REFERENCES_LINK_SOURCE = NO: Improves output. + VERBATIM_HEADERS = NO: The verbatim headers aren't very interesting. + PREDEFINED: Define XAPIAN_VISIBILITY_DEFAULT and XAPIAN_DEPRECATED(D)=D as doxygen seem reluctant to parse included files. + EXPAND_AS_DEFINED: Remove XAPIAN_VISIBILITY_DEFAULT and XAPIAN_DEPRECATED as these are now handled by PREDEFINED. + EXTERNAL_GROUPS = NO, GROUP_GRAPHS = NO: We don't use groups so disable these. Sun Jul 26 10:14:32 GMT 2009 Olly Betts <olly@survex.com> * docs/doxygen_api.conf.in,docs/doxygen_source.conf.in: Change @PACKAGE@ to @PACKAGE_NAME@ and @VERSION@ to @PACKAGE_VERSION@ to reflect modern autotools usage. Fri Jul 24 16:23:31 GMT 2009 Olly Betts <olly@survex.com> * configure.ac,docs/: Rename doxygen configuration files from "_conf" to ".conf". Rename doxygen_full.conf to doxygen_source.conf, etc. Fri Jul 24 16:02:39 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Update to note that doxygen is now installed in-tree, and that we now use 1.5.9 for 1.1.x snapshots and releases, and that graphviz ">1.8.10" is now required. Fri Jul 24 16:01:31 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/: Documentation comment tweaks. Fri Jul 24 15:34:03 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Fix @param names to match parameter names. Fri Jul 24 15:33:12 GMT 2009 Olly Betts <olly@survex.com> * docs/doxygen_api_conf.in: Major doxygen config overhaul: + STRIP_FROM_PATH: Strip include from header names. + JAVADOC_AUTOBRIEF: Re-enable as we have lost most of our brief descriptions - we can fix up the few cases where this caused problems with an explicit @brief. + BUILTIN_STL_SUPPORT: Enable for slightly STL-related stuff. + DISTRIBUTE_GROUP_DOC: Enable so group members get the group's comments. + EXTRACT_STATIC, EXTRACT_LOCAL_CLASSES: Disable since we don't need these. + SHOW_NAMESPACES: Disable to hide the "Xapian" namespace. + QUIET: Enable for less doxygen chatter. + WARN_LOGFILE: Unset so warnings go to stderr to encourage us to actually fix them! + EXCLUDE: Hide base.h, deprecated.h, errordispatch.h, visibility.h. + HTML_DYNAMIC_SECTIONS: Enable so the class diagrams are hidden by default but can be "expanded" to be viewed. + COLLABORATION_GRAPH: Disable as these don't ever seem useful in the API docs. + TEMPLATE_RELATIONS: Disable as we don't want to see template instantiations in inheritance diagrams. + DOT_MULTI_TARGETS: Enable for faster dot processsing. Needs graphviz ">1.8.10", but that's really old now. Fri Jul 24 13:00:24 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-compact.cc: Fix to work with the metainfo key stored in the latest format of chert databases. Fri Jul 24 10:32:32 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_values.cc: Avoid doing pointless work by trying to delete non-existent lists of values when we're just adding documents. Fri Jul 24 05:32:34 GMT 2009 Olly Betts <olly@survex.com> * docs/doxygen_api_conf.in: Update with "doxygen -u" using doxygen 1.5.9 and then trim trailing whitespace. Thu Jul 23 04:10:49 GMT 2009 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update for 1.1.2. Wed Jul 22 13:58:13 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_databasereplicator.cc, backends/flint/flint_databasereplicator.cc: Prefer += to + for building up strings. Wed Jul 22 09:01:57 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_databasereplicator.cc, backends/flint/flint_databasereplicator.cc: Check result of ::open() for -1, and throw an error immediately if this happens, rather than leaving it to a later call to fail with EBADF. Should make it easier to diagnose problems, since we'll now see the reason that the open() failed. Tue Jul 21 17:08:41 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog and sync with 1.0.14. Mon Jul 20 04:48:32 GMT 2009 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Fix build with assertions enabled. Mon Jul 20 04:47:33 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/query.h: Update doc comment - OP_SYNONYM can take any number of subqueries too. Sun Jul 19 17:26:53 GMT 2009 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,include/xapian/query.h,matcher/localmatch.cc: Remove wqf member of Xapian::Query::Internal and store the wqf in the parameter member instead. (ticket#280) Sun Jul 19 17:24:02 GMT 2009 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,include/xapian/query.h: Remove unused method Xapian::Query::Internal::swap(). Sun Jul 19 16:22:01 GMT 2009 Olly Betts <olly@survex.com> * api/valuerangeproc.cc,include/xapian/queryparser.h: Move support for a prefix/suffix from NumberValueRangeProcessor to StringValueRangeProcessor, and change NumberValueRangeProcessor and DateValueRangeProcessor to inherit from StringValueRangeProcessor so all three now support a prefix/suffix. (ticket#220) * tests/queryparsertest.cc: Add test coverage for new features. Sun Jul 19 16:05:04 GMT 2009 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Reenable tests which require the inmemory backend to be enabled by fixing typo XAPIAN_HAS_BACKEND_INMEMORY -> XAPIAN_HAS_INMEMORY_BACKEND. Sun Jul 19 14:56:16 GMT 2009 Olly Betts <olly@survex.com> * api/documentvaluelist.cc: Use str() instead of om_tostring(). Sun Jul 19 14:42:15 GMT 2009 Olly Betts <olly@survex.com> * matcher/msetcmp.cc: Eliminate two more relocations. Sun Jul 19 13:46:08 GMT 2009 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Factor out "is_distributable()" function. Sun Jul 19 11:11:06 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: Fix "dummy" -> "dummy[]" so the code here is valid C - we only preprocess it, but since the reason for having the dummy stuff at all is to avoid problems with "smart" preprocessors which moan if the code isn't valid, this seems worth fixing. Sun Jul 19 10:21:41 GMT 2009 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Backport the previous change to flint. Sun Jul 19 08:04:48 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc: Instead of dynamically building a std::list of the leafnames we need to replicate, just list them in a compact format in the source file. Results in smaller code and should be faster and smaller at runtime. Sat Jul 18 17:12:58 GMT 2009 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Merge among table entries which are substrings of others in the same among. We could be much cleverer and merge between amongs, and allow overlaps, etc, but this get us a nice size reduction for a small amount of effort, so it'll do for now. Sat Jul 18 15:42:48 GMT 2009 Olly Betts <olly@survex.com> * languages/compiler/generator.c,languages/steminternal.cc, languages/steminternal.h: Change how snowball generates the data used by among - instead of using pointers to the strings in struct among, store an offset into a constant pool, as this reduces the number of relocations from 5001 to 2706, which should decrease the time taken by the dynamic linker when loading the library. This also results in slightly smaller code. Fri Jul 17 16:32:02 GMT 2009 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Check output of Stem::get_description() for each supported language. Thu Jul 16 04:34:57 GMT 2009 Olly Betts <olly@survex.com> * common/fileutils.h: Remove unnecessary XAPIAN_VISIBILITY_DEFAULT (ticket#63). * common/fileutils.cc,common/fileutils.h: Make isabspath() static since it is only used by other functions in the same file. Thu Jul 16 04:13:48 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix comment typo. Thu Jul 16 03:59:49 GMT 2009 Olly Betts <olly@survex.com> * api/sorter.cc,include/xapian/sorter.h: MultiValueSorter doesn't need an explicit dtor - it's enough that Sorter has a virtual dtor. Thu Jul 16 03:29:41 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS: Add Andreas Flöter for the Solaris package howto. Wed Jul 15 14:37:19 GMT 2009 Olly Betts <olly@survex.com> * matcher/rset.cc: Just pass pointer as AutoPtr constructor parameter rather constructing a temporary AutoPtr and invoking the assignment operator. Wed Jul 15 03:43:35 GMT 2009 Olly Betts <olly@survex.com> * common/omenquireinternal.h: "class" -> "object" in doc comment. Wed Jul 15 03:38:55 GMT 2009 Olly Betts <olly@survex.com> * tests/runtest.in: Raise the fd limit from 32 to 64 - multi_flint tests need more than 32 fds. Tue Jul 14 20:32:52 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/omenquire.cc,common/omenquireinternal.h: Revert r13052, which changed Enquire::Internal::weight not to be mutable. Expand the comment explaining why it's mutable. Tue Jul 14 18:40:43 GMT 2009 Olly Betts <olly@survex.com> * tests/runtest.in: Use "ulimit -n" where available to limit the number of available file descriptors to 32 so we catch file descriptor leaks sooner. Tue Jul 14 16:53:48 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/omenquire.cc,common/omenquireinternal.h: Tidy up mutable Weight member on Enquire::Internal class; make it non-mutable, and just make a temporary BM25Weight object if the weight member was NULL. Tue Jul 14 12:36:46 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * common/output.h: Display a "<NULL $CLASSNAME>" message in debug output when a pointer being output is NULL. Previously, NULL Query::Internal pointers were causing segfaults with debug logging for many tests (eg, emptyquery1). Sun Jul 12 14:56:18 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc: We no longer have a "value" table (values are now stored in the postlist and termlist tables) so fix comments not to refer to it. Sun Jul 12 14:47:17 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/: Add new ChertLazyTable class and subclass lazy tables from it so that we only need to implement the common differences once. Sun Jul 12 13:41:27 GMT 2009 Olly Betts <olly@survex.com> * api/replication.cc: No point using join_paths() to append a literal leaf name, especially as we already don't elsewhere. Sun Jul 12 12:01:25 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix memory leak accidentally introduced in r13005. Sun Jul 12 11:40:26 GMT 2009 Olly Betts <olly@survex.com> * tests/api_opsynonym.cc: Remove debug "cout" calls. Sun Jul 12 09:43:06 GMT 2009 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Don't just skip counting subqueries which are the RHS of OP_AND_NOT - skip based on factor == 0.0, which means we get the RHS of OP_FILTER too. * tests/api_percentages.cc: Add topercent4 as a regression test for this. Sun Jul 12 09:32:11 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/andmaybepostlist.cc: Add DEBUGCALL macro to count_matching_subqs() to fix build with assertions. Sun Jul 12 08:17:02 GMT 2009 Olly Betts <olly@survex.com> * matcher/remotesubmatch.cc: Percentage scaling for remote results is handled in a different way, so replace FIXME with comment noting this. Sun Jul 12 05:48:17 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix handling of percentage weights in various cases when we're searching multiple remote databases or a mix of local and remote databases. * tests/api_percentages.cc: Add regression test topercent3. Sat Jul 11 13:35:02 GMT 2009 Olly Betts <olly@survex.com> * matcher/mergepostlist.cc: Fix file description - MergePostList can merge any number of posting lists, not just two. Sat Jul 11 13:09:40 GMT 2009 Olly Betts <olly@survex.com> * matcher/extraweightpostlist.h: ExtraWeightPostList::read_position_list() and ExtraWeightPostList::open_position_list() aren't used, so remove them. Sat Jul 11 12:54:30 GMT 2009 Olly Betts <olly@survex.com> * api/leafpostlist.cc,api/postlist.cc,common/leafpostlist.h, common/postlist.h,common/submatch.h,matcher/,tests/api_anydb.cc, tests/api_opsynonym.cc: Count how many leaf subqueries match for the document with the highest weight when calculating the percentage weights, instead of using the termlist of that document (ticket#363). Also fixes XOR with a SYNONYM subquery which could achieve 100% weight before. Fri Jul 10 08:02:18 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/chert/chert_spelling.cc, backends/flint/flint_spelling.cc,common/database.h: Deal with the case of get_spelling_suggestion() with an empty or single character word up front. Thu Jul 09 16:20:47 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc: In clean_up(), only scan through the pid_to_fd array once, and disable the signal handler and call waitpid() for any remaining pids ourselves, rather than calling sleep(1) and rechecking. This should avoid pointless delays when the child hasn't exited when we first check. Thu Jul 09 12:40:52 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/: BackendManager has a virtual dtor, so we don't need empty virtual dtors in its subclasses. Thu Jul 09 11:40:20 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc, tests/harness/backendmanager_remotetcp.h: Move the BackendManagerRemoteTcp ctor back into the header now it is trivial again. Wed Jul 08 16:56:53 GMT 2009 Olly Betts <olly@survex.com> * backends/flint/flint_io.h: Backport change to use F_FULLSYNC from chert. Wed Jul 08 15:30:22 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Check that we can find the valgrind/memcheck.h header as well as the valgrind binary. Wed Jul 08 14:30:53 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc: Sort out the clash between two different patches to fix leaking file descriptors. Also, call clean_up() from the destructor to correctly handle the case when a BackendManagerRemoteTcp is created by a test case directly. Wed Jul 08 14:17:43 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/: Rename "posttest()" method to "clean_up()" - I keep thinking "power-on self-test". Wed Jul 08 05:08:09 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_io.h: Use F_FULLFSYNC where available (Mac OS X currently). (ticket#288) Tue Jul 07 09:31:35 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: If default_op is OP_NEAR or OP_PHRASE then make the default window size (9 + no_of_terms), like it would be for an explicit NEAR or PHRASE without an explicit window size. (ticket#254) * tests/queryparsertest.cc: Add feature tests for this. Tue Jul 07 07:18:15 GMT 2009 Olly Betts <olly@survex.com> * HACKING,api/,backends/chert/,backends/flint/, backends/remote/remote-database.cc,common/,languages/steminternal.h, net/,queryparser/queryparser.cc,tests/harness/,tests/internaltest.cc, tests/perftest/freemem.cc,tests/perftest/perftest_randomidx.cc, tests/stemtest.cc,unicode/utf8itor.cc: Update to use C++ forms for ISO C standard headers (ticket#330). Tue Jul 07 05:30:58 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.h: Use "omassert.h" not <omassert.h>. Mon Jul 06 04:30:59 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS: Add Rolf Köhling for fixes for Omega on MS Windows. Sun Jul 05 16:28:54 GMT 2009 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_database.h: Backport the lazyupdate changes from chert to flint. Sun Jul 05 13:12:53 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc,tests/harness/backendmanager.h: Move virtual destructor definition out of the header. Sun Jul 05 04:09:04 GMT 2009 Olly Betts <olly@survex.com> * common/documentterm.h: Add const to OmDocumentTerm::get_wdf() method (ticket#139). Sat Jul 04 16:45:43 GMT 2009 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 6757" - no functional changes in generated code for our parser. Sat Jul 04 16:39:41 GMT 2009 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 6756" - no functional changes in generated code for our parser. Sat Jul 04 16:36:59 GMT 2009 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Merge upstream "Check-in Number: 6755" - no functional changes in generated code for our parser. Sat Jul 04 16:29:24 GMT 2009 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 6754" - no changes in generated code for our parser. Sat Jul 04 16:24:45 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 6749" - only changes are in code which isn't compiled due to preprocessor conditionals. Sat Jul 04 16:02:08 GMT 2009 Olly Betts <olly@survex.com> * api/editdistance.cc: Add const to is_transposed() method (ticket#139). Sat Jul 04 15:19:55 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_lock.h,backends/flint/flint_lock.h: Make ChertLock::operator bool() and FlintLock::operator bool() const (ticket#139). Sat Jul 04 14:56:42 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_database.h, backends/flint/flint_database.cc,backends/flint/flint_database.h: Make ChertDatabase::get_changeset_revisions() and FlintDatabase::get_changeset_revisions() const (ticket#139). Sat Jul 04 14:10:55 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_changesetapplier.h, backends/flint/flint_changesetapplier.h: "the the" -> "the" in comments. Sat Jul 04 14:06:39 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_btreebase.cc,backends/chert/chert_btreebase.h: Make ChertTable_base::do_unpack_uint() just a static non-class function. Use string::+= to build up the error string in it, and mark the error case as rare. Wed Jul 01 04:12:22 GMT 2009 Olly Betts <olly@survex.com> * docs/spelling.rst: Omega now supports spelling correction. Mon Jun 29 08:16:13 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Reset the begin and end strings before trying the next ValueRangeProcessor in case they've been modified. * tests/queryparsertest.cc: Add regression tests and general coverage for this area. Mon Jun 29 06:33:45 GMT 2009 Olly Betts <olly@survex.com> * backends/remote/net_postlist.cc: Remove unused '#include "serialise-double.h"'. Mon Jun 29 06:32:52 GMT 2009 Olly Betts <olly@survex.com> * backends/remote/net_postlist.cc: Build up description string using +=. Sun Jun 28 14:50:56 GMT 2009 Olly Betts <olly@survex.com> * backends/remote/net_postlist.h: Fix internal doc comment. Sat Jun 27 13:35:45 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: -Wshadow produces false positives with GCC 4.0, so only enable it for >= 4.1 since we enable -Werror for maintainer-mode builds for GCC >= 4.0. Sat Jun 27 13:30:22 GMT 2009 Olly Betts <olly@survex.com> * backends/multi/multi_termlist.cc,backends/multi/multi_termlist.h: Fix return type of MultiTermList::positionlist_count() to be Xapian::termcount rather than Xapian::termpos. Sat Jun 27 12:50:17 GMT 2009 Olly Betts <olly@survex.com> * backends/flint/flint_alltermslist.h,backends/flint/flint_postlist.cc, backends/flint/flint_postlist.h: Fix to use Xapian::doccount for termfreq in flint too. Sat Jun 27 12:28:47 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_alltermslist.cc: Prefer 'str.resize(0)' to 'str = ""'. Sat Jun 27 12:16:17 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_alltermslist.h,backends/chert/chert_postlist.h: Fix to use Xapian::doccount for termfreq in more places. Sat Jun 27 11:51:25 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc: Change so add_item_to_block() can always assume there is another contiguous space in the block (previously it could assume that there was enough space, but not that it was necessarily contiguous). Sat Jun 27 09:55:19 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/postingsource.h,include/xapian/weight.h: Add note that PostingSource and Weight objects returned by clone() and unserialise() methods will be deallocated with "delete". Sat Jun 27 07:26:41 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_postlist.cc: Adjust use of Xapian::doccount vs Xapian::termcount for consistency so things work better when these types aren't identical. Sat Jun 27 01:48:39 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/matchspy.cc: Remove unused matchspy.cc - there is an updated version of this on the matchspy branch; having an old copy here is just likely to lead to confusion. Fri Jun 26 05:42:26 GMT 2009 Olly Betts <olly@survex.com> * api/documentvaluelist.cc,api/documentvaluelist.h: Change DocumentValueList::skip_to()'s parameter type from Xapian::valueno to Xapian::docid to agree with the type in ValueList::skip_to(). This is a latent bug - it doesn't matter provided the types are actually the same (ticket#385). Thu Jun 25 11:20:07 GMT 2009 Olly Betts <olly@survex.com> * matcher/phrasepostlist.cc: Use Xapian::termcount(1) instead of 1u so that the code doesn't assume Xapian::termcount is unsigned int. Sun Jun 21 12:09:55 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_check.cc,backends/chert/chert_table.cc, backends/chert/chert_table.h: Replace duplicated macros with a single version in chert_table.h, and use DIR_START instead of a literal 11 in that header. * backends/chert/chert_table.h: Add more assertions. Use BYTES_PER_BLOCK_NUMBER instead of a literal 4 in more places. Sat Jun 20 07:38:56 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: Fix misspelling in comment (neccessary -> necessary). Sat Jun 20 07:38:06 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc: Don't reconstruct Item(p, c) when we already have it as a local variable. Wed Jun 17 01:23:16 GMT 2009 Olly Betts <olly@survex.com> * common/str.cc: Fix potential warning in signed version of str(). Tue Jun 16 11:20:43 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc,tests/harness/testsuite.h: Factor out writing out what was sent to tout and clearing it into a new method test_driver::write_and_clear_tout(). * tests/harness/testsuite.cc: Prefer string() to "". When sending a colour change code, do it before the preceding space when that allows string literals to be merged. Always give the full exception message (was conditional on --verbose). Tidy up output for reporting the various different exception types and other failures to give more consistent output. Tue Jun 16 11:01:14 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/scalability.cc: Increase the number of tests until the first sample takes more than 0.001 seconds to avoid trying to base calculations on a length of time we probably can't reliably measure to start with. Tue Jun 16 10:58:34 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/cputimer.cc: Don't bother trying to count CPU time for child processes, as it's hard to ensure that the child processes have finished with the current framework, and just counting time in the current process checks most things (the local backend tests cover much of the work done by the server part of the remote backend). Tue Jun 16 02:15:31 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/cputimer.h: Don't need '#include <ctime>'. Tue Jun 16 02:10:29 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/cputimer.h: Fix naming of header multiple-inclusion protection macro. Sun Jun 14 14:11:51 GMT 2009 Olly Betts <olly@survex.com> * tests/internaltest.cc: In testcase serialisequery1, use a helper function rather than wasting effort building a std::list with all the testcases in just so we can iterate over them. Sun Jun 14 13:52:22 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: <iomanip> isn't used so remove it. Sun Jun 14 13:30:34 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/testsuite.h: Fix comment typo. Sat Jun 13 06:30:57 GMT 2009 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Fix reversed conditional in last but one change. Sat Jun 13 04:42:27 GMT 2009 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Use TEST_EQUAL_DOUBLE() when comparing get_max_attained() to allow for rounding differences. Sat Jun 13 04:39:49 GMT 2009 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Use "startswith()" rather than comparing an extracted substring. Fri Jun 12 12:45:37 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc: Fix the mechanism for closing the file descriptors used by child xapian-tcpsrv processes once they exit, which has been broken ever since it was added nearly two years ago (sigh). Fixes fd leaking which caused "apitest -bremotetcp" to fail on OpenBSD (ticket#382). Fri Jun 12 03:37:13 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc: Fix comment typo. Wed Jun 10 04:29:43 GMT 2009 Olly Betts <olly@survex.com> * common/serialise-double.cc,common/serialise-double.h: Relicense as MIT/X rather than GPLv2+ as these are low level routines which other projects may find useful. Tue Jun 09 14:34:20 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update for 1.1.1. Tue May 26 14:55:05 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Update version to 1.1.1. Tue May 26 14:49:23 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Lick into shape. Mon May 25 11:55:59 GMT 2009 Olly Betts <olly@survex.com> * PLATFORMS: Sync with changes in 1.0 version. Mon May 25 10:57:59 GMT 2009 Olly Betts <olly@survex.com> * common/weightinternal.h: Remove declarations of unimplemented constructors (ticket#379). Sat May 23 16:06:05 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from 1.0.13 and ChangeLog. Sat May 23 15:35:39 GMT 2009 Olly Betts <olly@survex.com> * HACKING: ValueIterator is now a PIMPL class like TermIterator, etc. Sat May 23 05:24:33 GMT 2009 Olly Betts <olly@survex.com> * docs/postingsource.rst: Since we are sticking with an recursive maxweight recalculation for now, add back a warning about the potential overhead from calling set_maxweight() very frequently, with an informed estimate for the worst case overhead. Fri May 22 11:44:04 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc: We can trivially make Database::metadata_keys_begin() work for inmemory in the special case when there are no keys, so do so. * tests/api_metadata.cc: Add test coverage for this case. Fri May 22 09:31:07 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Only bother passing -Winit-self for GCC >= 4.3, as it doesn't do anything for GCC <= 4.4 anyway. Fri May 22 07:00:29 GMT 2009 Olly Betts <olly@survex.com> * backends/flint/flint_version.h: Initialise filename using + not +=. Fix odd whitespace in recent change. Fri May 22 05:27:56 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS: Update. Fri May 22 05:24:19 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc,common/multimatch.h,matcher/multimatch.cc: Replace MultiMatch::reduce_maxweight_by() with a call to MultiMatch::recalc_maxweight() and inline the later from the header. Fri May 22 05:16:03 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Enable more GCC warnings - "-Woverloaded-virtual" for all versions, "-Wstrict-null-sentinel" for 4.0+, "-Wlogical-op -Wmissing-declarations" for 4.3+. Notably "-Wmissing-declarations" caught that consistency2 wasn't being run. Fri May 22 05:15:08 GMT 2009 Olly Betts <olly@survex.com> * languages/steminternal.cc,languages/steminternal.h: Inline simple Stem::Internal::Internal() constructor from header. Thu May 21 16:23:26 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc: Make on_SIGCHLD() static. Thu May 21 16:22:56 GMT 2009 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Fix testcase consistency2 to actually be run. Thu May 21 16:22:30 GMT 2009 Olly Betts <olly@survex.com> * tests/api_transdb.cc: Add missing '#include "api_transdb.h"'. Thu May 21 12:58:22 GMT 2009 Olly Betts <olly@survex.com> * backends/databasereplicator.cc: Don't bother to check if the database path works in stat() and is a directory - the case we want to be fast is when the database does exist, and if path + "/iamflint" (or "iamchert") exists, then the check for path being a directory is redundant. If the path isn't a flint or chert database then a generic message including the path seems sufficient. Thu May 21 09:17:15 GMT 2009 Olly Betts <olly@survex.com> * backends/databasereplicator.cc,common/databasereplicator.h: Inline empty DatabaseReplicator default ctor. Thu May 21 08:25:24 GMT 2009 Olly Betts <olly@survex.com> * api/replication.cc: Cache the result of get_replica_path() when we are going to need it again. Thu May 21 08:09:51 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_databasereplicator.h, backends/flint/flint_databasereplicator.h: Correct parent class name in doxygen comments. Thu May 21 08:08:06 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Update list of Debian/Ubuntu packages needed to for a development environment. Thu May 21 06:28:00 GMT 2009 Olly Betts <olly@survex.com> * backends/databasereplicator.cc: Fix nested comment start introduced by the previous commit. Thu May 21 06:21:39 GMT 2009 Olly Betts <olly@survex.com> * api/replication.cc,backends/databasereplicator.cc, common/databasereplicator.h: We only keep one pointer to a DatabaseReplicator, so use AutoPtr not RefCntPtr. Thu May 21 05:00:17 GMT 2009 Olly Betts <olly@survex.com> * api/replication.cc: Fix SEGV when calling get_description() on a default constructed DatabaseReplica. * tests/generate-api_generated: Test DatabaseReplica (regression test for the above bug) and SerialisationContext. Fix @file in the generated api_generated.cc. Update (C) dates in generated api_generated.cc. Thu May 21 03:58:59 GMT 2009 Olly Betts <olly@survex.com> * tests/generate-api_generated: Call get_description() on the default constructed object to make sure that works (and doesn't try to dereference NULL, or fail some assertion, etc). All currently checked classes are fine - this is to avoid future regressions or such problems with new classes. Wed May 20 14:19:54 GMT 2009 Olly Betts <olly@survex.com> * net/tcpclient.cc,net/tcpserver.cc: Fix xapian-tcpsrv --interface option to work on MacOS X (ticket#373). Wed May 20 13:58:09 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc,api/sortable-serialise.cc,backends/chert/, backends/flint/,backends/remote/remote-database.cc, common/serialise-double.cc,common/str.cc, matcher/exactphrasepostlist.cc,net/remoteconnection.cc, net/remoteserver.cc: Replace C-style casts. Add a few missing casts. Tweak types used in a few places. Wed May 20 13:55:41 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_databasereplicator.cc, backends/flint/flint_databasereplicator.cc: Need <cstdio> for rename(). Replace C-style cast to off_t. Wed May 20 13:53:25 GMT 2009 Olly Betts <olly@survex.com> * api/editdistance.cc: Make edist_state class uncopyable and unassignable. Initialise maxdist in the initialiser list. Wed May 20 11:19:53 GMT 2009 Olly Betts <olly@survex.com> * languages/steminternal.cc,languages/steminternal.h: Change C-style casts to C++-style ones. Wed May 20 10:40:03 GMT 2009 Olly Betts <olly@survex.com> * common/str.h: Add static_cast<char>() to suppress bogus MSVC warning (ticket#377). Tue May 19 15:26:39 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/,backends/flint/,common/rset.h,tests/api_anydb.cc: Pass std::string by const reference rather than value (except in a few cases where we need a modifiable copy anyway) as benchmarking shows this is better even with GCC's reference counted std::string implementation (ticket#140). * tests/api_anydb.cc: Use string() instead of "" for an empty string. Tue May 19 08:47:55 GMT 2009 Olly Betts <olly@survex.com> * api/omquery.cc,include/xapian/query.h: Make Xapian::Query::MatchAll and Xapian::Query::MatchNothing const since they're immutable. All the public methods are const, so this should be completely API compatible. Tue May 19 06:25:26 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Document what passing maxitems=0 to get_mset() does. Mon May 18 13:32:49 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/database.h: Correct doxygen comments for user metadata functions: get_metadata() can't throw UnimplementedError but set_metadata() can. * include/xapian/database.h,tests/api_metadata.cc: Document that metadata_keys_begin() returns an end iterator if the backend doesn't support metadata, and add test coverage for this case. * backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h,include/xapian/database.h, tests/api_metadata.cc,tests/harness/testrunner.cc: The test harness didn't have the inmemory backend flagged as supporting user-specified metadata, and so it's perhaps not a great surprise to discover that inmemory doesn't support it fully - you can't iterate over metadata keys, but instead get an empty iteration. Fix this to at least throw UnimplementedError and document that this is the situation. Mon May 18 11:29:10 GMT 2009 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/api_metadata.cc,tests/api_wrdb.cc: Split out tests of user metadata into their own file. Sun May 17 12:52:41 GMT 2009 Olly Betts <olly@survex.com> * matcher/: Add a new constructor to MultiAndPostList for the "decay from OrPostList or AndMaybePostList" case which takes the already know max weights to avoid having to refetch them. Also, we know that the subpostlists of OrPostList are ordered a particular way so can avoid needing to check the order in that case. Sat May 16 05:02:35 GMT 2009 Olly Betts <olly@survex.com> * backends/flint/flint_version.cc: Fix comment - this file doesn't use memcpy() now. Thu May 14 04:38:47 GMT 2009 Olly Betts <olly@survex.com> * tests/api_percentages.cc: Remove superfluous explicit std:: qualifiers. Thu May 14 04:36:29 GMT 2009 Olly Betts <olly@survex.com> * tests/api_percentages.cc: Update for the PostingSource get_maxweight() API changes. Tue May 12 14:50:57 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Rename "max_weight" variable to "max_possible" since that is what it is referred to via the MSet API. Tue May 12 12:38:11 GMT 2009 Olly Betts <olly@survex.com> * common/safeuuid.h: Fix length of buffer to lower case - it's 36, not sizeof(uuid_t). Tue May 12 02:49:34 GMT 2009 Olly Betts <olly@survex.com> * common/safeuuid.h,configure.ac: Older versions of libuuid don't have uuid_unparse_lower() so probe for it in configure, and if it isn't present provide an inline version in safeuuid.h (ticket#368). Mon May 11 13:07:47 GMT 2009 Olly Betts <olly@survex.com> * tests/api_postingsource.cc: In changemaxweightsource1, set descending docid order so that the matcher isn't able to terminate early after 4 documents just because weight == maxweight. Mon May 11 12:00:04 GMT 2009 Olly Betts <olly@survex.com> * net/serialise.cc: Build up string using +=. Mon May 11 11:39:49 GMT 2009 Olly Betts <olly@survex.com> * docs/postingsource.rst: Improve documentation for get_maxweight(). Mon May 11 11:26:02 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/postingsource.h: Initialise max_weight_ to 0 in the PostingSource constructor in case nobody else sets a value for it. Mon May 11 10:54:47 GMT 2009 Olly Betts <olly@survex.com> * tests/api_postingsource.cc: Fix class name in FAIL_TEST message. Mon May 11 10:41:31 GMT 2009 Olly Betts <olly@survex.com> * tests/api_postingsource.cc: Fix get_termfreq_min(), etc for ChangeMaxweightPostingSource to return 4 not 5. Mon May 11 09:03:14 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc: Remove assertion checking that set_maxweight() doesn't set a lower weight - it legitimately can when we call init() on a non-clone-able PostingSource for reuse. Mon May 11 03:12:00 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc,common/multimatch.h,docs/postingsource.rst, include/xapian/postingsource.h,matcher/externalpostlist.cc, matcher/externalpostlist.h,matcher/msetpostlist.cc, tests/api_postingsource.cc: Replace PostingSource's notify_new_maxweight() and virtual get_maxweight() mechanism with a non-virtual set_maxweight() and get_maxweight(). Currently the plumbing at the matcher end is much as before, but this API seems more obvious to the user and will allow the matcher to simply reduce its maximum weight value rather than having to recursively call recalc_maxweight() in response to this situation (ticket#340). Mon May 11 02:38:26 GMT 2009 Olly Betts <olly@survex.com> * docs/postingsource.rst: Fix a typo and reword to avoid "namespaced". Mon May 11 02:30:59 GMT 2009 Olly Betts <olly@survex.com> * matcher/msetpostlist.cc: Fix typo - MSetPostList::get_maxweight() with a sort ordered primarily by decreasing relevance was returning the docid for the maxweight! This could have led to wrong results when searching multiple databases with the remote backend, but probably usually didn't matter as with BM25 the weights are usually small (often all < 1) while docids are inevitably >= 1. No regression test for this (currently anyway). Sun May 10 07:49:11 GMT 2009 Olly Betts <olly@survex.com> * backends/dbfactory.cc: Fix comment typo. Sat May 09 14:15:38 GMT 2009 Olly Betts <olly@survex.com> * tests/api_collapse.cc,tests/perftest/runprocess.cc: Remove unused include of <iostream>. Sat May 09 14:03:34 GMT 2009 Olly Betts <olly@survex.com> * api/omquery.cc,matcher/andnotpostlist.cc, queryparser/termgenerator_internal.cc,tests/api_wrdb.cc: Remove unused inclusions of <algorithm>. Sat May 09 13:47:54 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Remove unused inclusion of <queue>. Sat May 09 13:38:02 GMT 2009 Olly Betts <olly@survex.com> * api/omquery.cc,api/omqueryinternal.cc,api/replication.cc, backends/alltermslist.cc,backends/chert/chert_table.cc, backends/flint/flint_table.cc,bin/xapian-compact.cc, common/bitstream.cc,common/inmemory_positionlist.h, common/positionlist.h,matcher/localmatch.h,matcher/queryoptimiser.cc, matcher/queryoptimiser.h,tests/api_wrdb.cc: Remove unused inclusions of <list>, <map>, <string>, and <vector>. Sat May 09 12:48:18 GMT 2009 Olly Betts <olly@survex.com> * api/documentvaluelist.cc,matcher/queryoptimiser.cc: Remove unused '#include "autoptr.h"'. Sat May 09 09:41:03 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,queryparser/termgenerator.cc, weight/bm25weight.cc,weight/boolweight.cc,weight/tradweight.cc: Remove unneeded explicit "std::" qualifiers. Sat May 09 09:35:32 GMT 2009 Olly Betts <olly@survex.com> * weight/weight.cc: Doesn't use '#include "autoptr.h"'. Fri May 08 19:59:00 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Change source file which autoconf looks for, since I've just removed the old one that it looked for! Fri May 08 19:28:31 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/andmaybepostlist.cc,matcher/branchpostlist.h: Call check() instead of skip_to() on the optional branch of AND_MAYBE. This can be much faster if that branch is a posting source. Fri May 08 18:31:01 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/: Remove AndPostList, in favour of MultiAndPostList. AndPostList was only used as a decay product (by AndMaybePostList and OrPostList); for most queries, MultiAnd has pretty much indistinguishable performance as far as I can test; however, it uses check() when possible, which makes performance much better with searches involving external posting sources. Removal also reduces the amount of code cluttering up the caches during a search, so may help speed things up in some other situations. Fri May 08 16:00:42 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc,backends/inmemory/inmemory_database.cc, examples/delve.cc,include/xapian/postingsource.h: Don't store an end iterator in a variable just to compare against it - it's now definitely more efficient to compare against the return value of the relevant _end() method. Fri May 08 15:44:07 GMT 2009 Olly Betts <olly@survex.com> * api/valueiterator.cc,include/xapian/database.h, include/xapian/document.h,include/xapian/valueiterator.h: Create a new proxy object class ValueIteratorEnd_ which is returned instead of ValueIterator(NULL) for end iterators. This class will be converted to ValueIterator(NULL) if assigned to or passed as ValueIterator, but operator== and operator!= are overloaded to handle it directly which means that the compiler no longer calls ~ValueIterator for each call which returns ValueIterator(NULL). Fri May 08 15:42:12 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_values.cc: Swap two statements for cosmetic reasons. Fri May 08 14:42:49 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc,include/xapian/postingsource.h: It is actually better not to store the end iterator in a member variable - if we explicitly call db.postlist_end(string()) then that just gets optimised to PostingSource(NULL). Fri May 08 14:02:21 GMT 2009 Olly Betts <olly@survex.com> * tests/api_postingsource.cc: We've decided that "going back" with skip_to() or check() should have unspecified behaviour, so stop testing how this case behaves! Fri May 08 12:37:33 GMT 2009 Olly Betts <olly@survex.com> * common/Makefile.mk: Ship common/win32_uuid.cc and common/win32_uuid.h (ticket#367). Fri May 08 12:30:45 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/cputimer.cc: Fix typo in the "ftime" case. Fri May 08 10:25:37 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/andpostlist.cc: Fix a swap which was the wrong way round, so that we call next() on the subpostlist with the lower frequency, rather than the one with the higher frequency. This should improve performance in many situations, and I don't think it's likely to slow anything down. It also fixes some cases where optimisation of an AND_MAYBE to an AND results in a slower search, such that search times went down when more hits were requested. Fri May 08 08:48:14 GMT 2009 Olly Betts <olly@survex.com> * matcher/externalpostlist.cc: ExternalPostList::get_doclength() should never get called, so make it "Assert(false)". Fri May 08 05:16:50 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc: Another string concatention. Fri May 08 05:04:55 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc: Use str() instead of om_tostring() and build up strings using +=. Fri May 08 04:37:05 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc: "using namespace std;". Thu May 07 16:46:14 GMT 2009 Olly Betts <olly@survex.com> * matcher/: Check for collection_size == 0 up front so we don't need to repeatedly special case it when implementing get_termfreq_est_using_stats(). Thu May 07 16:28:07 GMT 2009 Olly Betts <olly@survex.com> * matcher/xorpostlist.cc: Fix typo bug - collection_size should be rset_size. This would have lead to a floating point division by zero when XOR was used under a synonym without an RSet, but that would just give infinity, and so just give a bogus value for the rel_termfreq rather than aborting instantly. Thu May 07 16:19:52 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Update auto_ptr information. Mention constructor syntax as an alternative replacement for some C-style casts. Thu May 07 15:25:00 GMT 2009 Olly Betts <olly@survex.com> * matcher/localmatch.cc,matcher/localmatch.h: Don't rebuild the termname -> termfreq and weight map for every subdatabase - just do it once for the first one. Also eliminate a copy of this map. This should speed up searches a little, especially those over multiple databases. Thu May 07 15:01:24 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc,api/omqueryinternal.cc, backends/remote/remote-database.cc,common/autoptr.h, expand/esetinternal.cc,matcher/localmatch.cc: Make our AutoPtr just a macro wrapper for std::auto_ptr. Since std::auto_ptr doesn't allow assignment from a raw pointer, this requires changing such assignments into calls to auto_ptr::reset(). Thu May 07 14:19:22 GMT 2009 Olly Betts <olly@survex.com> * api/leafpostlist.cc: Add an assertion that the term in the stats. Add "using namespace std;" so we can avoid explicit std:: qualifiers. Thu May 07 13:26:21 GMT 2009 Olly Betts <olly@survex.com> * matcher/: Eliminate EmptySubMatch - we can just store NULL for a submatch which has failed but which ErrorHandler told us to continue without. Thu May 07 12:52:24 GMT 2009 Olly Betts <olly@survex.com> * common/database.h: Update internal documentation comment - Database::Internal::open_post_list() can no longer return EmptyPostList as it is no longer a subclass of LeafPostList. Thu May 07 12:13:50 GMT 2009 Olly Betts <olly@survex.com> * api/leafpostlist.cc,backends/chert/chert_postlist.cc, backends/chert/chert_postlist.h, backends/flint/flint_alldocspostlist.h, backends/flint/flint_postlist.cc,backends/flint/flint_postlist.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h,backends/remote/net_postlist.h, common/contiguousalldocspostlist.h,common/leafpostlist.h: Squash TermBasedLeafPostList into LeafPostList since there are no longer any other subclasses. Thu May 07 11:49:06 GMT 2009 Olly Betts <olly@survex.com> * api/Makefile.mk,api/emptypostlist.cc,common/emptypostlist.h: Reimplement EmptyPostList from scratch, and inherit from PostList not LeafPostList. Thu May 07 11:19:34 GMT 2009 Olly Betts <olly@survex.com> * matcher/mergepostlist.cc: Add explicit '#include "omassert.h"'. Thu May 07 10:28:33 GMT 2009 Olly Betts <olly@survex.com> * backends/multi/multi_postlist.cc,backends/multi/multi_postlist.h: Subclass MultiPostList directly from PostList instead of from LeafPostList. This gets rid of two unused data members per MultiPostList in exchange for having to define 5 extra "never called" methods, but 4 of these just tailcall. Thu May 07 10:26:39 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_alltermslist.cc, backends/inmemory/inmemory_database.cc: Fix assertion failures and a SEGV in an assertion test. Thu May 07 09:27:35 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_alltermslist.cc, backends/inmemory/inmemory_alltermslist.h, backends/inmemory/inmemory_database.cc: Keep a dummy entry in InMemoryDatabase::postlists so that the implementation of InMemoryAllTermsList doesn't need a "started" flag, and we can use an InMemoryPostList on the dummy item for the "term not in database" case. Thu May 07 05:45:48 GMT 2009 Olly Betts <olly@survex.com> * tests/api_backend.cc: Add tset coverage for alldocs iterator on an empty database. * backends/inmemory/inmemory_database.cc: No need to special case an empty database as InMemoryAllDocsPostList handles this case correctly. Thu May 07 05:08:07 GMT 2009 Olly Betts <olly@survex.com> * weight/weightinternal.cc: Use str() instead of om_tostring(), += to build up strings, and fix a cut-and-paste comment error. Wed May 06 13:03:40 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * net/serialise.cc: Better serialisation - don't include the reltermfreqs if the rset_size is 0, since they'll all be 0 too. Wed May 06 12:13:52 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * common/remoteprotocol.h,net/serialise.cc: When serialising stats, serialise the termfreq and reltermfreq together, rather than in separate lists. Results in a smaller serialised form, and matches the in-memory structure now used better. Fixes ticket #362. This is an incompatible remote protocol change, however, so bump the major version to 32. Wed May 06 04:02:43 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Check termfreqandwts for MatchAll queries before we check terms from the document so that the "terminate early once we've seen all the query terms" test when checking the document terms can fire in this case. Wed May 06 03:26:05 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Remove incorrect comment accidentally committed in r12623. Wed May 06 03:18:42 GMT 2009 Olly Betts <olly@survex.com> * tests/api_opsynonym.cc: Fix comment typo. Wed May 06 01:35:14 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/api_postingsource.cc: changemaxweightsource1 uses ChangeMaxweightPostingSource, which doesn't work with multi or remote, so mark the test accordingly. Wed May 06 01:11:25 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/postingsource.h,api/postingsource.cc: Add notify_new_maxweight() method to be called by subclasses when they want to notify the matcher that their maxweight has changed significantly. Add ExternalPostList as a friend and add register_externalpl() to be used to associate the ExternalPostList with the PostingSource. * matcher/externalpostlist.cc,matcher/externalpostlist.h, matcher/queryoptimiser.cc: Register the externalpostlist with the PostingSource it's iterating through. Add ExternalPostList::notify_new_maxweight() method for PostingSource::notify_new_maxweight() to call. * docs/postingsource.rst: Document notify_new_maxweight() and the reasons for calling it. * tests/api_postingsource.cc: Test the handling of maxweight. Tue May 05 19:37:56 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am,tests/api_db.cc,tests/api_postingsource.cc: Split the tests of PostingSource subclasses out of api_db.cc; I'm about to add some more, and bloating api_db.cc further is unworkable. Tue May 05 18:23:02 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * docs/postingsource.rst,include/xapian/postingsource.h: Documentation updates to describe way that init() is called again if a PostingSource is reused. Fixes #352. Tue May 05 16:05:48 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/api_opsynonym.cc: Test that the top document has a weight of 100%, which is should do for all the current examples here. Tue May 05 16:03:29 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: Add logging to keep track of the weights being added to percent_scale. If the percent_scale total is 0.0, which can happen if the top document only matches synonym terms, force percent_scale to 1.0, to avoid an assertion failure. Tue May 05 13:41:57 GMT 2009 Olly Betts <olly@survex.com> * common/str.h: Fix swapped parameters in string ctor in code for str(bool) overload which lead to us returning 48 or 49 \x01 characters instead of "0" or "1". * tests/internaltest.cc: Add regression test strbool1. * AUTHORS: Thank Peter Kelm for helping us spot this. Tue May 05 12:54:36 GMT 2009 Richard Boulton <richard@lemurconsulting.com> Merge from opsynonym branch: * queryparser/queryparser.lemony: Use OP_SYNONYM instead of OP_OR for synonyms, wildcards, and partial queries. * tests/queryparsertest.cc: Test use of OP_SYNONYM by query parser. Tue May 05 12:18:06 GMT 2009 Richard Boulton <richard@lemurconsulting.com> Merge from opsynonym branch: * api/omqueryinternal.cc,include/xapian/query.h,matcher/, weight/weight.cc: Implement a new OP_SYNONYM query operator, which behaves similarly to OP_OR, but produces weights for the combined set of terms as if they were a single term (by merging their wdfs). * tests/Makefile.am,tests/api_opsynonym.cc: Add tests for the OP_SYNONYM operator. * include/xapian/weight.h: Add a Weight::init_() method for synonym, since there isn't a single term involved. Also, requires a method get_sumpart_needs_wdf_() to check if wdfs are needed, since the synonym weight calculation requires fetching the doclength, which is potentially expensive, in order that the wdf returned can be clamped to a sane range. Mark the weighting methods which require WDF with need_stat(WDF). Tue May 05 12:04:33 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * common/remoteprotocol.h: Add comment to remind use to do the FIXME in serialise_stats() when next breaking remote protocol compatibility. Tue May 05 10:53:11 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * common/output.h: Add output function for TermFreqs. Tue May 05 08:23:28 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/leafpostlist.cc,api/postlist.cc,common/emptypostlist.h, common/leafpostlist.h,common/postlist.h,matcher/: Add PostList::get_termfreq_est_using_stats() method, with default implementation which raises an InvalidOperationError. Implement this for those postlists which it makes sense for. Also, implement get_wdf() for postlists used by the matcher for implementing various query operators. This isn't currently used or tested on trunk, but is used for the opsynonym branch. Tue May 05 07:37:04 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * common/rset.h,common/weightinternal.h,matcher/rset.cc, net/serialise.cc,weight/weightinternal.cc: Store termfreqs and reltermfreqs in a Weight::Internal in a single map; introduce a new struct (TermFreqs) for the values stored here. Remove now unused Xapian::TermFreqMap typedef. Tue May 05 07:19:19 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_postlist.cc,backends/chert/chert_postlist.h, backends/flint/flint_alldocspostlist.h, backends/flint/flint_postlist.cc,backends/flint/flint_postlist.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h,backends/remote/net_postlist.h, common/contiguousalldocspostlist.h,common/leafpostlist.h: Introduce TermBasedLeafPostList, as a new common base class for postlists which are derived from a term in the query (including alldocs postlists which are derived from the empty term). Tue May 05 06:47:11 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/,backends/flint/flint_modifiedpostlist.cc, backends/flint/flint_postlist.cc,backends/flint/flint_postlist.h: Rename "tname" parameters to LeafPostList subclasses to "term". * backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h: Rename term parameter of type InMemoryTerm to imterm ready for opsynonym branch changes. Tue May 05 04:51:28 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_alldocspostlist.h,matcher/multimatch.cc: Remove unused '#include "leafpostlist.h"'. Tue May 05 02:15:35 GMT 2009 Olly Betts <olly@survex.com> * common/contiguousalldocspostlist.h: Wrap comment. Tue May 05 02:04:02 GMT 2009 Olly Betts <olly@survex.com> * common/Makefile.mk: Move str.h into alphabetical order. Mon May 04 12:04:16 GMT 2009 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Rename qp_stem_scale1 to qp_scale1 as it has no connection with stemming. Factor out a helper function to do the timing comparison, and update time_query_parse() to use CPUTimer rather than OmTime. Tweak the "fudge factor" up from 2 to 2.15 to allow for non-equal ticks due to rounding of an interval which isn't an exact multiple of 1/CLK_TCK (ticket#308). Mon May 04 12:02:32 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/cputimer.cc: If we don't have rusage() or times(), but do have ftime(), use that instead of time() so we at least get subsecond resolution (as we do currently with OmTime). Mon May 04 03:31:51 GMT 2009 Olly Betts <olly@survex.com> * docs/queryparser.html: Add examples of using a prefix on a phrase or subexpression. Sun May 03 14:58:42 GMT 2009 Olly Betts <olly@survex.com> * tests/api_valuestats.cc: Testcase valuestats4 requires transactions, so indicate that and remove the explicit SKIP for inmemory. Sun May 03 12:34:30 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * common/rset.h,common/weightinternal.h,matcher/rset.cc, weight/weightinternal.cc: Typedef std::map<string, Xapian::doccount> as TermFreqMap; tidies up code a little here, but mainly useful for keeping code tidy on branches/opsynonym. Sun May 03 11:49:08 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * common/multimatch.h: Remove unused include of <map> Sun May 03 05:23:32 GMT 2009 Olly Betts <olly@survex.com> * tests/api_db.cc: Protect tests which assume Flint is enabled with #ifdef XAPIAN_HAS_FLINT_BACKEND. Sat May 02 15:49:46 GMT 2009 Olly Betts <olly@survex.com> * configure.ac,tests/Makefile.am,tests/api_scalability.cc, tests/api_wrdb.cc,tests/harness/: Add new "test_scalability()" helper function which checks that an operation scales in CPU time used in the desired way. Use this to reimplement bigoaddvalue (and rename to bigoaddvalue1 for consistency). Fri May 01 04:58:41 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-check.cc: For chert, check value stats are the correct format and that the streamed values are consistent with their stats (ticket#277). Fri May 01 04:53:22 GMT 2009 Olly Betts <olly@survex.com> * api/omdocument.cc: Don't ever store empty values explicitly, which eliminates an inconsistently handled corner case which was causing the "value frequency" to be wrong from at least one point of view. This is consistent with how user metadata works too. * include/xapian/document.h: Explicitly document this behaviour. * tests/api_valuestats.cc: Add regression test for the above change. Fri May 01 03:11:25 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Chert doesn't store termlist entries for documents without terms, which resulted in us reporting an error when we found document ids in the doclength "postlist" which were greater than any with an entry in the termlist. Instead compare these entries against db.get_last_docid() if we are checking a whole db and able to call it. If not, suppress this check. Thu Apr 30 15:02:41 GMT 2009 Olly Betts <olly@survex.com> * common/valuestats.h: Fix comment typo. Tue Apr 28 14:01:41 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc: Use <cstdlib> rather than <stdlib.h> which means that there's an overload for abs(long). Tue Apr 28 13:17:48 GMT 2009 Olly Betts <olly@survex.com> * HACKING,INSTALL,configure.ac,include/xapian/deprecated.h, include/xapian/version_h.cc,tests/harness/testsuite.cc: Actually, let's require GCC 3.1 as doing so eliminates some preprocessor conditionals which we aren't able to test regularly as we don't have easy access to such old GCC versions. GCC 3.1 is nearly 7 years old now, and GCC3 didn't get widespread use until later versions anyway. Tue Apr 28 12:47:11 GMT 2009 Olly Betts <olly@survex.com> * common/debuglog.h: Add missing initialisation of uncaught_exception in a couple of places. Tue Apr 28 11:24:02 GMT 2009 Olly Betts <olly@survex.com> * INSTALL: Note that Xapian 1.0.x should build with GCC 2.95.3. Tue Apr 28 11:20:51 GMT 2009 Olly Betts <olly@survex.com> * HACKING,INSTALL,backends/chert/chert_database.h, backends/flint/flint_database.h, backends/inmemory/inmemory_database.h,configure.ac, include/xapian/version_h.cc,tests/harness/testsuite.cc: Drop support for GCC 2.95.3 - we now require at least 3.0. Mon Apr 27 12:20:25 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/weight.h: Add "need_stat(WDF);" for Weight subclasses which use wdf. Currently this is never actually checked by trunk (so a regression test for this omission isn't really feasible), but the opsynonym branch uses it. Sun Apr 26 14:47:57 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * weight/bm25weight.cc: Add call debugging. Sun Apr 26 10:57:48 GMT 2009 Olly Betts <olly@survex.com> * tests/api_db.cc: stubdb6 is "inmemory" not "flint || chert || multi". Sun Apr 26 10:51:50 GMT 2009 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Fix to build with the inmemory backend disabled. Sun Apr 26 07:28:47 GMT 2009 Olly Betts <olly@survex.com> * tests/generate-api_generated: Fix so generated api_generated.cc builds with the inmemory backend disabled. Sun Apr 26 06:30:57 GMT 2009 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Fix to build with the inmemory backend disabled. Sat Apr 25 01:10:49 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Fix to build with the flint and/or chert backends disabled. Sat Apr 25 01:06:00 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix to build with the remote backend disabled. Sat Apr 25 00:57:41 GMT 2009 Olly Betts <olly@survex.com> * backends/dbfactory.cc: Fix to build with the inmemory backend disabled. Fri Apr 24 22:59:34 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * weight/weight.cc: Add debug logging to Weight::init_() methods. * net/remoteserver.cc: Fix bounds returned by Weight::Internal object in get_mset() in remote submatches, by setting the database used to read the bounds. Fri Apr 24 13:14:50 GMT 2009 Olly Betts <olly@survex.com> * matcher/branchpostlist.cc,matcher/branchpostlist.h: Remove BranchPostList::get_wdf() which isn't currently used, and isn't wanted on branches/opsynonym. Fri Apr 24 12:46:18 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Apply doxygen comment improvement from branches/opsynonym. Fri Apr 24 12:31:15 GMT 2009 Olly Betts <olly@survex.com> * matcher/localmatch.cc: Apply assertion and debug logging improvements from branches/opsynonym. Thu Apr 23 06:16:05 GMT 2009 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,docs/Makefile.am,tests/Makefile.am: Fix things up so that in a bootstrapped SVN tree, automatic regeneration of autotools-generated files uses the in-tree versions of the autotools. Thu Apr 23 04:37:31 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS: Add another bug reporter. Thu Apr 23 04:26:25 GMT 2009 Olly Betts <olly@survex.com> * Makefile.am,languages/Makefile.mk,queryparser/Makefile.mk, tests/Makefile.am,tests/perftest/Makefile.mk: Don't remove any built sources in "make clean" even under --make-maintainer-mode as that breaks switching a tree away from maintainer-mode with: make distclean;./configure Thu Apr 23 03:51:03 GMT 2009 Olly Betts <olly@survex.com> * Makefile.am: Use "--coverage" instead of "-fprofile-arcs -ftest-coverage" in the test coverage build. Wed Apr 22 14:32:56 GMT 2009 Olly Betts <olly@survex.com> * PLATFORMS: Sync with 1.0.12; add entry for atreus for 1.1.0. Wed Apr 22 14:27:36 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS: Add more bug reporters and patch contributors. Remove a duplicate entry. Wed Apr 22 13:58:31 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Set release date. Wed Apr 22 13:36:04 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update, perhaps even ready for release now. Wed Apr 22 11:00:50 GMT 2009 Olly Betts <olly@survex.com> * NEWS: More work. Tue Apr 21 17:59:13 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * docs/queryparser.html: Add closing <code> tag. Tue Apr 21 11:05:17 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * NEWS: A few minor corrections and adjustments. Tue Apr 21 07:49:49 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/weight.h: Don't use the term "term" to talk about the extra weight component as it already has a strong meaning in Xapian. Tue Apr 21 07:16:18 GMT 2009 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 6451". Tue Apr 21 06:54:00 GMT 2009 Olly Betts <olly@survex.com> * matcher/queryoptimiser.h: '#include <map>' isn't required. Tue Apr 21 06:45:14 GMT 2009 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc,matcher/queryoptimiser.h: Eliminate QueryOptimiser::do_leaf() as it is only called from one place and if only two lines of code. Mon Apr 20 14:10:08 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog and start to lick into shape for a release. Mon Apr 20 13:06:17 GMT 2009 Olly Betts <olly@survex.com> * HACKING: XAPIAN_DEBUG_LOG=- send output to stderr, not stdout. Mon Apr 20 11:53:07 GMT 2009 Olly Betts <olly@survex.com> * common/output.h: Add new macro XAPIAN_OUTPUT_FUNCTION_PTR and use it so that debug logging of Xapian::Query::Internal* doesn't just report the pointer value. Mon Apr 20 06:39:42 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Apply debug logging change lifted from branches/opsynonym. Mon Apr 20 05:22:24 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/testutils.cc: Add FIXME comments about use of internal macro from a different header. Mon Apr 20 05:08:34 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Update release checklist. Sun Apr 19 14:50:08 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/api_nodb.cc: Stick the reciprocal of "factor" into a volatile double temporary variable to stop compiler using extended precision calculation on x86, and coming up an answer of exactly 1.0 when multiplying by "factor" again. Sun Apr 19 14:26:09 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Sync with 1.0.12 and update from ChangeLog. Sun Apr 19 12:12:19 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS: Update for 1.0 branch. Thu Apr 16 12:08:24 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_values.cc: When copying the tail of a value chunk because did > last_allowed_did, we don't need to check if the docids being copied are < did, since they should all be <= last_allowed_did, so just assert the latter. Tue Apr 14 20:32:41 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/api_wrdb.cc: patch from james to move logging in modifyvalues1 so it reports the values it's rewriting correctly, rather than the empty string. Also display the expected value in tout when checking the value. Also, pull the setting of the random seed out to the start of the function, and set it to 7 (which exposed the bug fixed with the previous commit, which that the old value of 42 didn't, on my machine, at least). Tue Apr 14 18:12:08 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_values.cc: When we move the reader to a new chunk, write any trailing values left in it to the stream first. Thu Apr 09 05:01:04 GMT 2009 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Document the child process used for locking which exec-s "cat" (ticket #258). Thu Apr 09 03:49:13 GMT 2009 Olly Betts <olly@survex.com> * bin/: Use C++ forms of C headers in bin (ticket#330). Thu Apr 09 03:47:42 GMT 2009 Olly Betts <olly@survex.com> * examples/: Use C++ forms of C headers in examples (ticket#330). Mon Apr 06 06:35:03 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Fix documentation comment typos. Mon Apr 06 06:03:45 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS: Add Muayyad Alsadi for reporting #355. Mon Apr 06 05:59:38 GMT 2009 Olly Betts <olly@survex.com> * tests/api_unicode.cc: Note which characters in the new unicodepredicate1 testcase are new in Unicode 5.1.0. Sun Apr 05 14:24:47 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Add NON_SPACING_MARK to is_wordchar() for better tokenisation of Arabic, for example. (ticket#355) * tests/queryparsertest.cc: Add test that this has the desired effect. * tests/api_unicode.cc: Add feature tests of Unicode::is_wordchar(), Unicode::is_currency(), and Unicode::is_whitespace(). Sat Apr 04 05:41:47 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Wed Apr 01 00:18:02 GMT 2009 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Fix totlen overflow test. Refactor lightly to reduce indenting. Tue Mar 31 17:01:20 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-compact.cc: Handle databases which contain no documents (and hence have no METAINFO item), but do contain some metadata (so the postlist table is not empty). Fixes ticket #356. Tue Mar 31 13:49:01 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc,include/xapian/postingsource.h: Change return types of clone() and unserialise() methods to be of the subclass returned; this is a bit more flexible, but the main reason is to be consistent: previously, the clone() methods returned the subclass, but the unserialise() methods returned PostingSource. Tue Mar 31 12:10:03 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/postingsource.h: Add comment on PostingSource base class listing it as an experimental feature. * docs/deprecation.rst: Describe what "experimental" features are, and why replication and posting sources are currently experimental. Tue Mar 31 11:34:25 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/matchspy.h: Removed currently unused header to stop doxygen from generating documentation for it. Tue Mar 31 11:27:54 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * docs/replication.rst,include/xapian/replication.h: Note that replication is currently "experimental". Fri Mar 27 14:29:05 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_spelling.cc,backends/chert/chert_spelling.h, backends/flint/flint_spelling.cc,backends/flint/flint_spelling.h: Fix WritableDatabase::remove_spelling() to not be very broken in several ways. * tests/api_spelling.cc: Add test coverage for it. Fri Mar 27 10:43:34 GMT 2009 Olly Betts <olly@survex.com> * xapian-config.in: Add @LIBRARY_VERSION_SUFFIX@ to -lxapian too. Fri Mar 27 08:22:39 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/postingsource.h: Update documentation comments to refer to init() instead of the old name of reset(). Fri Mar 27 08:10:48 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Document the lcov make targets. Fri Mar 27 07:44:59 GMT 2009 Olly Betts <olly@survex.com> * Makefile.am: Add two new targets to assist generating test coverage reports with lcov: coverage-reconfigure which reruns configure in the source tree and coverage-check which runs "make check" and generates an HTML report in a directory called "lcov". Thu Mar 26 23:59:36 GMT 2009 Olly Betts <olly@survex.com> * Makefile.am,m4-macros/xapian.m4,xapian-core.spec.in: Rename xapian.m4 to xapian-1.1.m4 to avoid clash with Xapian 1.0.x. Both xapian.m4 and xapian-1.1.m4 will contain XO_LIB_XAPIAN, but aclocal copes and only copies one of them. We need to avoid incompatible changes to XO_LIB_XAPIAN though. Thu Mar 26 14:29:54 GMT 2009 Olly Betts <olly@survex.com> * configure.ac,include/Makefile.mk,xapian-config.in: Install headers in $prefix/include/xapian-1.1/ so they don't clash with 1.0. Thu Mar 26 13:46:06 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Default program suffix to -1.1 if not specified. If you really want no suffix, "./configure --program-suffix=" will achieve that. Thu Mar 26 12:49:10 GMT 2009 Olly Betts <olly@survex.com> * Makefile.am,api/Makefile.mk,backends/Makefile.mk, backends/chert/Makefile.mk,backends/flint/Makefile.mk, backends/inmemory/Makefile.mk,backends/multi/Makefile.mk, backends/remote/Makefile.mk,bin/Makefile.mk,common/Makefile.mk, configure.ac,examples/Makefile.mk,expand/Makefile.mk, languages/Makefile.mk,matcher/Makefile.mk,net/Makefile.mk, queryparser/Makefile.mk,tests/Makefile.am,tests/perftest/Makefile.mk, unicode/Makefile.mk,weight/Makefile.mk,xapian-config.in, xapian-core.spec.in: Change library name to libxapian-1.1 as a first step towards allowing parallel installation with 1.0.x. Thu Mar 26 06:53:27 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Thu Mar 26 06:50:44 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_values.h: Fix class name in file doc comment. Thu Mar 26 00:13:48 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/flint/flint_database.cc: Call ChertDatabase::close() rather than repeating its code. Wed Mar 25 17:39:01 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_database.cc,backends/flint/flint_database.cc: Release the lock when the database is closed (either due to a call to close() or due to a serious error during modifications). Should fix ticket #354. Wed Mar 25 16:42:58 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/harness/testutils.cc: Use TEST_EQUAL_DOUBLE_ for comparison of weights in mset, to fix test failures due to floating point differences. (Currently only manifests in some tests in the matchspy branch, but it's a generally useful fix.) Wed Mar 25 11:44:36 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Only put ANSI_CXXFLAGS in CXXFLAGS for the duration of configure - put it in AM_CXXFLAGS for substituting so that the user can safely override CXXFLAGS at make-time: "make CXXFLAGS=-Os" Wed Mar 25 10:50:44 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Fix comment typo. Wed Mar 25 04:48:50 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/chert/chert_database.h, backends/flint/flint_database.h, backends/inmemory/inmemory_database.h, backends/remote/remote-database.cc,common/const_database_wrapper.h, common/database.h: Drop the default value of the lazy parameter to Database::open_document() - it's better to force us to consider whether a new call should be lazy or not. * common/database.h: Better description of lazy parameter. * backends/remote/remote-database.cc: Update comment about when lazy is set to true. Wed Mar 25 04:20:34 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc: Don't "return false" when the return type is a pointer! Tue Mar 24 17:42:22 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/database.cc: Open documents lazily in collect_document(). Tue Mar 24 11:10:57 GMT 2009 Olly Betts <olly@survex.com> * backends/multi/multi_postlist.cc,backends/multi/multi_postlist.h: MultiPostList::read_position_list() is never used, so remove it. Tue Mar 24 10:04:03 GMT 2009 Olly Betts <olly@survex.com> * backends/multi/multi_postlist.cc: Remove "[" and "]" from the output of MultiPostList::get_description(). Remove excess "," from output of MultiPostList::get_description(). * tests/api_db.cc: Add some test coverage of PostingIterator::get_wdf() and PostingIterator::get_description(). Tue Mar 24 10:00:23 GMT 2009 Olly Betts <olly@survex.com> * api/ompostlistiterator.cc: Remove "[" and "]" from Xapian::PostingIterator::get_description() as they don't add anything useful. Tue Mar 24 09:57:39 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc: Add a space in the output of InMemoryPostList::get_description() and InMemoryAllDocsPostList::get_description(). Tue Mar 24 08:52:46 GMT 2009 Olly Betts <olly@survex.com> * api/leafpostlist.cc,include/xapian/weight.h,tests/api_db.cc,weight/: Get rid of the virtual Weight::get_sumpart_needs_doclength() method - subclasses can call need_stat(DOC_LENGTH) in their constructor if they need doc lengths. Tue Mar 24 07:55:06 GMT 2009 Olly Betts <olly@survex.com> * weight/weight.cc: Check stats_needed for stats which aren't just a simple member variable lookup. Tue Mar 24 06:46:24 GMT 2009 Olly Betts <olly@survex.com> * backends/multi/multi_postlist.cc,backends/multi/multi_postlist.h: MultiPostList::get_termfreq() is never used, so remove it. Mon Mar 23 14:09:33 GMT 2009 Olly Betts <olly@survex.com> * tests/api_unicode.cc: Add test coverage for Xapian::Unicode::append_utf8(). Mon Mar 23 11:40:09 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc,docs/postingsource.rst, include/xapian/postingsource.h,matcher/externalpostlist.cc, tests/api_db.cc,tests/api_percentages.cc,tests/api_valuestream.cc: Rename PostingSource::reset() to PostingSource::init(). Mon Mar 23 11:01:02 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/testsuite.h: Fix documentation comment in line with the corrected --help output from "Sat Feb 21 08:18:21 GMT 2009". Mon Mar 23 02:24:44 GMT 2009 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Fix handling of empty and single subquery OP_NEAR and OP_PHRASE (was breaking queryparsertest). Sun Mar 22 10:23:37 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/flint/flint_database.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_positionlist.cc, common/inmemory_positionlist.h,docs/deprecation.rst, tests/api_posdb.cc: Change Database::positionlist_begin() not to throw exceptions if the term or document doesn't exist. Sun Mar 22 09:02:46 GMT 2009 Olly Betts <olly@survex.com> * docs/deprecation.rst: Fix typo. Sun Mar 22 07:54:30 GMT 2009 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,include/xapian/query.h,tests/api_nodb.cc, tests/api_query.cc: Fix valgrind errors on nearsubqueries1 (ticket#349) and enhance distribution of OP_NEAR/OP_PHRASE over non-leaf subqueries to work when there are multiple non-leaf subqueries (ticket#201). Extend nearsubqueries1 to test that trying to distribute OP_NEAR/OP_PHRASE over OP_NEAR/OP_PHRASE throws UnimplementedError. Fri Mar 20 08:43:20 GMT 2009 Olly Betts <olly@survex.com> * tests/api_query.cc: Add coverage for OP_AND with MatchNothing subquery (new testcase matchnothing1). Fri Mar 20 05:16:18 GMT 2009 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Fix assertions in previous check-in to verify the right condition. And return the simplified query in the case the assertion would have failed, as that means a small memory leak rather than unpredictable behaviour in a non-assertion build. Fri Mar 20 04:26:29 GMT 2009 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Check that end_construction() returns NULL when unserialising queries. Fri Mar 20 02:51:32 GMT 2009 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Fix return type of Xapian::Query::Internal::simplify_query() in a debug build. Since RETURN() isn't used by this function, this is only a cosmetic issue in the log file. Wed Mar 18 09:26:02 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/flint/flint_database.cc, docs/deprecation.rst: Move just returning an empty PositionIterator from Database::positionlist_begin() when the docs or term isn't present from 1.2.0 to 1.1.0 (has to be 1.1.0 or 1.3.0 really). Wed Mar 18 09:24:27 GMT 2009 Olly Betts <olly@survex.com> * docs/glossary.rst: Chert as the default backend in 1.2.0 is pretty much definite. Wed Mar 18 06:42:09 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Wed Mar 18 04:36:59 GMT 2009 Olly Betts <olly@survex.com> * xapian-core.spec.in: Tweak reference to "1.1.0 branch" to make more sense. Wed Mar 18 04:30:18 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc,include/xapian/postingsource.h: Remove the optional "max_weight_" parameter to ValuePostingSource's ctor. * include/xapian/postingsource.h: Correct a few errors in documentation comments. Wed Mar 18 04:29:22 GMT 2009 Olly Betts <olly@survex.com> * docs/postingsource.rst: "ctx" -> "context" in example code. Tue Mar 17 23:45:52 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_lock.cc,backends/flint/flint_lock.cc: If the locking attempt fails after we spawn the child process, call waitpid() in the parent process to avoid creating a zombie process. * AUTHORS: Add Jim Spath for reporting this. Tue Mar 17 23:34:16 GMT 2009 Olly Betts <olly@survex.com> * common/Makefile.mk,common/str.cc,common/str.h: New family of overloaded str() functions for converting types to std::string. The integer conversion are much faster than the existing om_tostring(). * common/omassert.h,common/omdebug.h: Use str() instead of om_tostring(). * common/utils.cc,common/utils.h: Make om_tostring() a macro wrapping str() for now to avoid introducing a lot of conflicts with other branches and unapplied patches. Tue Mar 17 21:28:54 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: Fix comment typo. Tue Mar 17 00:47:26 GMT 2009 Olly Betts <olly@survex.com> * common/utils.cc: Fix previous change to actually compile... Mon Mar 16 20:10:01 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * common/utils.cc,common/utils.h: Add om_tostring() version of unsigned long long int, to fix debug build on i686. Mon Mar 16 14:01:02 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Sync with 1.0.11. Fri Mar 13 06:59:29 GMT 2009 Olly Betts <olly@survex.com> * weight/weightinternal.cc: Need "utils.h" for om_tostring(). Fri Mar 13 05:47:24 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Note that QueryParser::FLAG_DEFAULT was new in 1.0.11. Fri Mar 13 04:36:19 GMT 2009 Olly Betts <olly@survex.com> * weight/weightinternal.cc: Implement Weight::Internal::get_description(). Thu Mar 12 11:57:02 GMT 2009 Olly Betts <olly@survex.com> * common/output.h: Fix last commit. Thu Mar 12 11:29:58 GMT 2009 Olly Betts <olly@survex.com> * common/output.h: Define operator<< for Xapian::Weight::Internal. Thu Mar 12 11:21:09 GMT 2009 Olly Betts <olly@survex.com> * net/remoteserver.cc: Fix overlooked case which was still sending REPLY_DOCLENGTH as a double. Wed Mar 11 22:56:59 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_valuelist.cc: Fix comment typo. Wed Mar 11 22:50:49 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/api_valuestream.cc: Enable valuestream3 for chert. * backends/chert/chert_valuelist.cc: Fix return value of ChertValueList::check() when it needs to move to a new chunk. Fixes valuestream3 test. Wed Mar 11 13:17:58 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/weight.h: Mark Weight::init_() methods as "@private @internal" for doxygen API docs. Wed Mar 11 13:13:32 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/weight.h,tests/api_db.cc,tests/api_nodb.cc, weight/bm25weight.cc,weight/boolweight.cc,weight/tradweight.cc: Make Weight::name() return std::string to allow sane wrapping with SWIG directors. Wed Mar 11 05:46:40 GMT 2009 Olly Betts <olly@survex.com> * api/,backends/chert/,backends/contiguousalldocspostlist.cc, backends/flint/,backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/multi/multi_postlist.cc,backends/multi/multi_postlist.h, backends/remote/,common/,docs/remote_protocol.html, include/xapian/database.h,include/xapian/postingiterator.h,matcher/, net/remoteserver.cc: Internally, pass around non-normalised document lengths as Xapian::termcount (unsigned integer) not Xapian::doclength (double). Wed Mar 11 05:25:40 GMT 2009 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Use TEST_REL; wrap long comment. Wed Mar 11 04:52:41 GMT 2009 Olly Betts <olly@survex.com> * Makefile.am,api/,backends/chert/,backends/database.cc, backends/flint/flint_database.cc,backends/flint/flint_database.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/remote/remote-database.cc,common/,docs/remote_protocol.html, examples/delve.cc,include/Makefile.mk,include/xapian.h, include/xapian/database.h,include/xapian/enquire.h, include/xapian/weight.h,matcher/,net/remoteserver.cc, net/serialise.cc,tests/,weight/: Reimplementation of weighting schemes which allows user subclasses to access the same stats which built in schemes can (bug#213) and which also can use lower and upper bounds of doclength and upper bounds on wdf to give a tighter maxweight, which should allow the matcher weight-based optimisations to be more effective. Chert now tracks doclength bounds and a global (rather than per term) bound on wdf. Other backends return much less good bounds, but these still lead to better maxweight bounds. Wed Mar 11 00:13:24 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/query.h: Rename parameter name from "ctx" to the more human-friendly "context". Tue Mar 10 14:03:33 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/testutils.cc: If mset_range_is_same_percents() fails because the percentages differ, report what they were to tout. Mon Mar 09 12:27:28 GMT 2009 Olly Betts <olly@survex.com> * matcher/localmatch.cc: Remove special case handling for empty string being returned from the TermIterator returned by Query::Internal::get_terms() as this should no longer happen. Mon Mar 09 12:09:21 GMT 2009 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Don't return "" from a TermIterator from Query::get_terms_begin() (happened when Query contained or was Query::MatchAll). * tests/Makefile.am,tests/api_query.cc: Add regression testcase queryterms1. Sun Mar 08 06:16:24 GMT 2009 Olly Betts <olly@survex.com> * tests/api_backend.cc: Add check that backends don't truncate total document length to 32 bits. Sun Mar 08 06:02:43 GMT 2009 Olly Betts <olly@survex.com> * tests/api_backend.cc: Disable lockfileumask1 on Cygwin and on OS/2. Sun Mar 08 05:36:54 GMT 2009 Olly Betts <olly@survex.com> * matcher/extraweightpostlist.h: Add forward declaration of Xapian::Weight since we use "Xapian::Weight *". Sun Mar 08 04:39:22 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_types.h,backends/flint/flint_types.h, bin/xapian-check-flint.cc,bin/xapian-check.cc,bin/xapian-compact.cc, common/Makefile.mk,common/database.h,common/internaltypes.h: Move totlen_t to a new header (internaltypes.h) to avoid needing to pull in database.h in when we want to use it. Also move int4 and uint4 there to avoid duplicated definitions and abstract out a new uint8 from defining totlen_t. Sun Mar 08 00:50:38 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.h,backends/chert/chert_types.h, backends/flint/flint_database.h,backends/flint/flint_types.h, bin/xapian-check-flint.cc,bin/xapian-check.cc,bin/xapian-compact.cc, common/database.h: Replace flint_totlen_t and chert_totlen_t with a common totlen_t. Sat Mar 07 08:35:27 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc: Wrap comment. Sat Mar 07 08:28:17 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.h: Fix file description ("multiple database" -> "inmemory database"). Sat Mar 07 08:25:11 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h: Store non-normalised document lengths as Xapian::termcount (unsigned int) rather than Xapian::doclength (double). Sat Mar 07 06:56:26 GMT 2009 Olly Betts <olly@survex.com> * common/database.h: Removed unused forward declarations of classes. Sat Mar 07 06:52:11 GMT 2009 Olly Betts <olly@survex.com> * common/database.h: Fix comment typo. Sat Mar 07 05:44:18 GMT 2009 Olly Betts <olly@survex.com> * net/serialise.cc: Fix top of file to actually be a doxygen comment. Tweak layout of definition of serialise_stats(). Sat Mar 07 05:40:58 GMT 2009 Olly Betts <olly@survex.com> * net/serialise.cc: c_str() -> data() since we don't need the nul-termination. Thu Mar 05 11:49:45 GMT 2009 Olly Betts <olly@survex.com> * tests/api_db.cc,tests/api_nodb.cc: Back out accidentally committed changes 2 commits ago. Thu Mar 05 10:28:07 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS: Add Frank J. Bruzzaniti for omindex patches. Thu Mar 05 10:20:38 GMT 2009 Olly Betts <olly@survex.com> * tests/Makefile.am: If both chert and flint are enabled, run remoteprog_chert and remotetcp_flint but not the other two combinations as they don't buy us any useful extra test coverage, but cost a lot of extra time per testrun. Thu Mar 05 07:51:52 GMT 2009 Olly Betts <olly@survex.com> * docs/deprecation.rst: Update to reflect that odd minor versions are now development series. Thu Mar 05 04:20:43 GMT 2009 Olly Betts <olly@survex.com> * common/serialisationcontextinternal.h: Fix forward declarations. Thu Mar 05 02:11:26 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc: "new_uuid" -> "sub_uuid" - it's not really new. Wed Mar 04 22:53:13 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,include/xapian/termgenerator.h: Make default empty std::string parameters use std::string() rather than "". Wed Mar 04 16:06:11 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc: If a sub-database of a multi-database has no uuid, we cannot return a uuid for the database, so return string() instead. * tests/api_anydb.cc: Test for this. Wed Mar 04 13:45:41 GMT 2009 Olly Betts <olly@survex.com> * HACKING: Conversion to C++ names for ISO C headers will happen in 1.1.x, not 1.1.0. Wed Mar 04 13:37:47 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/database.cc, backends/remote/remote-database.cc,common/database.h, include/xapian/database.h,net/remoteserver.cc,tests/api_anydb.cc: Return an empty string from Database::get_uuid() when then backend doesn't support UUIDs for consistency with flint when the database has no UUID. For a multiple database, build a string from the UUIDs of each subdatabase and return this as the UUID. If there are no subdatabases return the empty string. Wed Mar 04 11:34:04 GMT 2009 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remoteprotocol.h, common/remoteserver.h,docs/remote_protocol.html,net/remoteserver.cc: Change the remote server to always default to opening a Database and having the client request write access explicitly. This allows a single server to support multiple readers and one writer simultaneously. (bug#145) * net/remoteserver.cc,backends/remote/remote-database.cc: For constant empty string parameters, pass string() rather than "". * common/remoteserver.h: Remove unused '#include <map>'. Remove 'using namespace std;' and explicitly qualify the few bare mentions of 'string'. Wed Mar 04 07:24:39 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Sync with latest upstream lempar.c (only changes were in code which isn't in our version). Wed Mar 04 04:03:44 GMT 2009 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/api_percentages.cc, tests/testdata/apitest_sortconsist.txt: Apply regression test patch from bug#216. Wed Mar 04 03:41:49 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/enquire.h,matcher/multimatch.cc: Fix inconsistent percentage scores when sorting primarily by valuei, except when a MatchDecider is also being used; document this remaining problem case. (bug#216) Wed Mar 04 01:58:07 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: No need to initialise inflate_zstream->next_out and avail_out twice. Wed Mar 04 01:16:32 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: If deflateReset() or inflateReset() fails, just delete the zstream take the initialisation codepath - that should mean we recover if the stream does somehow end up in a bad state and is less code than throwing an exception. Annotate tests with rare() or usual() if they are for exceptional conditions. Fix bug introduced by the laziness patch which was adding an int error code to a const char * message. If deflateInit2() or inflateInit2() fails, delete and zero the stream so there's no risk of ending up wedged on a partly initialised stream. Tue Mar 03 15:32:53 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: Check for errors from deflateReset() and inflateReset() - there shouldn't be any if we've given them a valid stream, so just raise InternalError if we get one. Tue Mar 03 09:43:34 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Updated from ChangeLog. Tue Mar 03 05:52:16 GMT 2009 Olly Betts <olly@survex.com> * tests/api_closedb.cc: Fix testcase closedb2 not to create a test database named closedb3. Tue Mar 03 04:31:48 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Add QueryParser::FLAG_DEFAULT to make it easier to add flags to those set by default. * tests/queryparsertest.cc: Use FLAG_DEFAULT in a few places to serve as a feature test. Tue Mar 03 03:45:53 GMT 2009 Olly Betts <olly@survex.com> * docs/index.html: Add link to new serialisation topic document. * docs/serialisation.rst: Reword to avoid pluralising class names. Add link to postingsource document where we reference it. Tue Mar 03 03:35:29 GMT 2009 Olly Betts <olly@survex.com> * tests/runsrv.in: Use @top_builddir@ rather than relying on it coming from the environment. Tue Mar 03 03:05:47 GMT 2009 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Clear tout regularly in modifyvalues1 as otherwise the output builds up. Some string stream implementations get very inefficient with large strings, and also this seems to interact badly with valgrind's leak detection resulting in the test having to be rerun to check if it really leaks. Prefer tout.str(string()) to tout.str("") in existing uses too. Tue Mar 03 01:41:07 GMT 2009 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Update handling of shutting down the zstream objects in the dtor to match chert. Mon Mar 02 23:57:26 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_table.h,backends/flint/flint_table.h: Add documentation comments for the lazy allocation methods, and separate comments for the zstream members. Mon Mar 02 23:40:20 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * AUTHORS,backends/chert/chert_table.cc,backends/chert/chert_table.h, backends/flint/flint_table.cc,backends/flint/flint_table.h: Reduce allocation of zstreams by allocating them lazily, keeping them in the table objects, and reuse them, rather than allocating a new one each time it's wanted. Heavily based on a patch from Todd Lipcon. Apparently improves performance significantly on some systems; on my tests it has no perceptible negative impact, but it makes sense that it could help in some situations, so let's apply it. Fixes #325. Mon Mar 02 21:10:54 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Deprecate Stem_get_available_languages() from the python bindings. Mon Mar 02 19:56:20 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * common/Makefile.mk,common/const_database_wrapper.cc, common/const_database_wrapper.h,matcher/queryoptimiser.cc: Avoid const_cast() on Database::Internal, and exposing compiler-dependent behaviour if non-const methods are called, by adding a ConstDatabaseWrapper class, which is a subclass of Database::Internal, and also takes a Database::Internal as a parameter, to be wrapped. This class proxies all const methods to the wrapped Database::Internal, and raises InvalidOperationError on non-const methods. Fixes #332. Mon Mar 02 18:27:00 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/postingsource.h: Documentation comment improvements. Be explicit that Xapian will call reset() before various methods, and will call next, skip_to or check before at_end. * include/xapian/query.h: Wrap over-long line. Mon Mar 02 18:26:37 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * docs/Makefile.am,docs/serialisation.rst: Add topic document about serialisation. Mon Mar 02 17:58:28 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * docs/postingsource.rst: Update documentation with new design. Mon Mar 02 17:16:54 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/omqueryinternal.cc: Change serialisation of queries to use encode_length() rather than om_tostring(), for a more compact and easier to parse representation. Mon Mar 02 14:48:08 GMT 2009 Olly Betts <olly@survex.com> * backends/multi/multi_alltermslist.cc: Fix memory leak which was causing allterms6 to fail. Mon Mar 02 14:15:11 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/externalpostlist.cc: Fix memory leak in external source postlist (fixes fixedweightsource1). Mon Mar 02 13:41:29 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc: Don't set max_weight in ValueMapPostingSource::set_default_weight() - it's set in reset() before being used, and it's not been set here yet, so accessing it in std::max() was making valgrind complain. Mon Mar 02 12:23:01 GMT 2009 Olly Betts <olly@survex.com> * HACKING,configure.ac,tests/runtest.in: Don't use valgrind if it's < 3.3.0 as that's well over a year old and greatly simplifies the configure tests. Fix options passed to valgrind so that the testsuite harness actually reports problems detected by newer valgrind versions. Mon Mar 02 10:08:25 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/,bin/xapian-tcpsrv.cc,common/Makefile.mk,common/remoteserver.h, common/serialisationcontextinternal.h,include/Makefile.mk, include/xapian.h,include/xapian/enquire.h,include/xapian/query.h, include/xapian/serialisationcontext.h,net/remoteserver.cc, tests/api_serialise.cc,tests/internaltest.cc: Add SerialisationContext. This object is used to register PostingSource or Weight subclasses, so that they can be serialised and unserialised, and therefore used in remote searches. Add two-argument form for Query::unserialise() which takes a serialisation context. Fixes #206. Mon Mar 02 07:13:38 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/database.h: Further tweak to wording of documentation about close(). Mon Mar 02 07:05:54 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_table.cc,backends/chert/chert_table.h, backends/flint/flint_table.cc,backends/flint/flint_table.h: Factor out more "throw Xapian::DatabaseError("Database has been closed")" bits into separate functions: knocks another 20KB off the shared library size. Mon Mar 02 05:46:15 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Document how to get all matches from Enquire::get_mset(). Mon Mar 02 04:52:17 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_alltermslist.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h: Factor out "throw Xapian::DatabaseError("Database has been closed") into a separate function which knocks over 100KB off the shared library size (just over 0.5%). Mon Mar 02 03:57:17 GMT 2009 Olly Betts <olly@survex.com> * matcher/phrasepostlist.h: Remove blank line from end of file. Mon Mar 02 03:52:10 GMT 2009 Olly Betts <olly@survex.com> * api/replication.cc,common/msvc_posix_wrapper.cc: Cuddle braces for "if" and "switch". Mon Mar 02 03:32:32 GMT 2009 Olly Betts <olly@survex.com> * tests/api_closedb.cc: Use more conventional uppercase macro parameter names. Wrap long comments. Remove another unused exception name. Mon Mar 02 03:21:39 GMT 2009 Olly Betts <olly@survex.com> * api/replication.cc,backends/chert/chert_databasereplicator.cc, backends/flint/flint_databasereplicator.cc,net/remoteserver.cc, tests/api_closedb.cc,tests/perftest/perftest_matchdecider.cc: Don't name the exception being caught if we don't look at it. While GCC doesn't currently warn "unused variable" here, I'm fairly sure there are compilers which do. Mon Mar 02 03:20:52 GMT 2009 Olly Betts <olly@survex.com> * tests/perftest/perftest.cc: Wrap overlong comment. Mon Mar 02 03:17:48 GMT 2009 Olly Betts <olly@survex.com> * matcher/mergepostlist.cc: Change "do { ... } while (true);" to the more usual "while (true) { ... }". Sun Mar 01 19:41:16 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/api_closedb.cc: Merge closedb1 and closedb2 into a single test, and add lots more checking to it. Rename closedb3 to closedb2, to avoid leaving an odd gap. closedb1 needs more work, but is a useful start; committing now, rather than once I've done this work, since the old test is failing in buildbot, and I don't want to get in the way of snapshot tarballs. Sun Mar 01 19:27:38 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: Add a few guards to throw DatabaseError when the database has been permanently closed, rather than behaving as if the table is empty in that situation. Also, don't delete the resources in the table when Database::close() is called, because they may still be used to attempt to look up cached content. Sun Mar 01 18:22:52 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/database.h: Update documentation comments for reopen() and close() to document the API we're aiming for (ie, that closed databases either return the right result from cache, or raise a DatabaseError). Sun Mar 01 18:20:19 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/inmemory/inmemory_alltermslist.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h: Add a flag to mark whether the database is closed, and add guards before all methods which access content which goes away when the database is closed, to raise an exception if the database is closed. Sun Mar 01 18:20:03 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/inmemory/inmemory_document.h: Fix a typo in a comment. Sun Mar 01 17:49:26 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/mergepostlist.cc: Fix ticket #336 by calling recalc_maxweight when a sub-posting list finishes (other than after the last one, where there is no point in calling recalc_maxweight). Sun Mar 01 15:34:46 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/api_db.cc: Fix MyDontAskWeightPostingSource to use the database passed to reset(), rather than one passed to its constructor. This allows externalsource4 to be enabled for multidatabases. Sun Mar 01 13:07:37 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/postingsource.h: Add comments as discussed in ticket #295 to make clear that in a multi-database search, PostingSources are passed a single sub-database. Also mention that the returned value of clone() will be deallocated with delete. Sat Feb 28 08:21:33 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/chert/chert_database.cc, backends/chert/chert_database.h,backends/database.cc, backends/flint/flint_database.cc,backends/flint/flint_database.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/remote/remote-database.cc,common/,docs/remote_protocol.html, examples/copydatabase.cc,include/xapian/database.h, net/remoteserver.cc,tests/,tests/perftest/perftest_matchdecider.cc, tests/perftest/perftest_randomidx.cc,tests/queryparsertest.cc: Add WritableDatabase::commit() as a new, preferred alias for WritableDatabase::flush(). (bug#266) Fri Feb 27 01:59:31 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/database.h: WritableDatabase::flush() can't throw DatabaseLockError. WritableDatabase's ctor can throw at least DatabaseCorruptError or DatabaseLockError. Thu Feb 26 14:54:28 GMT 2009 Olly Betts <olly@survex.com> * tests/api_db.cc,tests/api_valuestream.cc: Update comments about why certain tests are disabled for certain backends. Enable valueweightsource5 for multi and valuemapsource2 for remote. Thu Feb 26 14:53:19 GMT 2009 Olly Betts <olly@survex.com> * matcher/multiandpostlist.cc: Use AssertRelParanoid() instead of AssertParanoid() so the values get reported if the assertion fails. Thu Feb 26 14:18:21 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/postingsource.h: Tweak word order. Thu Feb 26 12:48:16 GMT 2009 Olly Betts <olly@survex.com> * tests/api_db.cc: Make the PostingSource subclass ctors which are only called from clone() private to set a good example for users. Thu Feb 26 12:28:34 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/postingsource.h: Fix doc comment typo. Thu Feb 26 12:08:40 GMT 2009 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,include/xapian/query.h: Avoid copying Query::Internal objects needlessly when unserialising Query objects. Thu Feb 26 10:07:18 GMT 2009 Olly Betts <olly@survex.com> * tests/api_db.cc: A couple more uncollapsed tests. Thu Feb 26 08:28:42 GMT 2009 Olly Betts <olly@survex.com> * tests/api_anydb.cc,tests/api_wrdb.cc,tests/harness/testsuite.h, tests/perftest/perftest_matchdecider.cc,tests/queryparsertest.cc: Update to use new TEST_REL() macro. Remove old TEST_LESSER(), etc. Thu Feb 26 07:51:43 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix the new Enquire::get_uncollapsed_matches_lower_bound(), etc methods for certain cases. * tests/api_anydb.cc,tests/api_db.cc: Extend existing tests to also check Enquire::get_uncollapsed_matches_lower_bound(), etc. Thu Feb 26 07:49:23 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/Makefile.mk,tests/harness/testmacros.h, tests/harness/testsuite.h: Add new "TEST_REL" macro which can test for a condition using any relational operator and report a failure clearly yet concisely. Wed Feb 25 13:43:41 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix adjustment of the uncollapse bounds and estimate. Wed Feb 25 12:32:13 GMT 2009 Olly Betts <olly@survex.com> * configure.ac: GCC --version keeps changing format, and as a result we were currently getting "g++" as the version. So change to the (hopefully) more robust technique of using g++ -E to pull out __GNUC__ and __GNUC_MINOR__. Wed Feb 25 03:14:08 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc: Fix for compiling with Sun's compiler (untested as I no longer have access to it). Mon Feb 23 14:11:12 GMT 2009 Olly Betts <olly@survex.com> * docs/sorting.rst: Clarify meaning. Mon Feb 23 14:04:50 GMT 2009 Olly Betts <olly@survex.com> * docs/deprecation.rst,include/xapian/enquire.h,tests/: Rename the wrongly named "ascending" parameter of the set_sort_by*() methods of Enquire to "reverse" and deprecate the default value since defaulting to "reverse=true" is confusing. (bug#311) Mon Feb 23 01:24:23 GMT 2009 Olly Betts <olly@survex.com> * api/omenquire.cc,backends/remote/remote-database.cc, common/multimatch.h,common/omenquireinternal.h, common/remote-database.h,docs/collapsing.rst, docs/remote_protocol.html,include/xapian/enquire.h,matcher/, net/remoteserver.cc,net/serialise.cc,tests/Makefile.am, tests/api_collapse.cc: Replace the collapsing code in the matcher with a separate Collapser class. This new class can keep more than one document with each collapse key. Also track bounds and an estimate of the total number of matches if collapsing wasn't in use. Mon Feb 23 00:52:16 GMT 2009 Olly Betts <olly@survex.com> * api/omdatabase.cc: Fix AssertionError from apitest testcases fixedweightsource2 and valueweightsource1. Sun Feb 22 11:02:24 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_lock.cc,backends/flint/flint_lock.cc: Check for fork failing *after* we check if we're now the child process. Sun Feb 22 10:06:13 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_lock.cc,backends/flint/flint_lock.cc: Note explicitly why we now have the strange empty if (not) checking the return value of chdir("/") (it's actually newer glibc and _FORTIFY_SOURCE rather than newer GCC). Retry closing filehandle if it returns EINTR. Whitespace tweaks. Sat Feb 21 11:04:08 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_lock.cc,backends/flint/flint_lock.cc: Fix warning about ignoring result of chdir (with GCC 4.2) Sat Feb 21 08:18:21 GMT 2009 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Fix option usage in --help - "-x=foo" doesn't work - "-x foo" does. Sat Feb 21 07:37:01 GMT 2009 Olly Betts <olly@survex.com> * tests/perftest/perftest_randomidx.cc: Use two argument version of rand_int() where applicable. Reserve length of result string in gen_word(). Fix truncated comment. Sat Feb 21 04:27:45 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_lock.cc,backends/flint/flint_lock.cc: In the locking child process, before we exec /bin/cat change directory to / so that we don't block unmounting of any partitions and close any open file descriptors (apart from those we're using) so that if the files are closed by our parent and deleted the disk space gets released right away. Sat Feb 21 02:06:56 GMT 2009 Olly Betts <olly@survex.com> * api/postingsource.cc,include/xapian/postingsource.h: Pass string as const string &; drop trailing "_" from parameters which don't clash with member variables; fix brace formatting on a couple of for loops. Fri Feb 20 14:32:46 GMT 2009 Olly Betts <olly@survex.com> * tests/api_closedb.cc: Fix testcase closedb3 not to create a test database named closedb2. Fri Feb 20 01:38:32 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Documentation comment improvements. Wed Feb 18 06:48:43 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Use AssertRel rather than Assert with an inequality so that we see the arguments if the assertion fails. Wed Feb 18 06:39:33 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: "items" -> "rsetitems" to distinguish from items used elsewhere for the MSet items. Wed Feb 18 06:20:38 GMT 2009 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Drop superfluous "std::". Wed Feb 18 05:46:59 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Tue Feb 17 23:53:54 GMT 2009 Olly Betts <olly@survex.com> * include/xapian/document.h,include/xapian/query.h: "xapian" -> "Xapian". Tue Feb 17 15:02:49 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * include/xapian/document.h,include/xapian/query.h: Soften comment warning about changes in serialised format between xapian versions; note that the format won't change unless the remote database protocol has changed. Tue Feb 17 14:59:29 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/omquery.cc: Return std::string() rather than "" to encourage compilers to use special empty string representation. Mon Feb 16 14:27:28 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc: Change the name() methods of the built-in posting sources to return a full name (starting with Xapian:: and ending with PostingSource), as documented in the API documentation comments. Mon Feb 16 13:32:39 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/omqueryinternal.cc,common/remoteserver.h, matcher/externalpostlist.cc,matcher/externalpostlist.h, matcher/queryoptimiser.cc,net/remoteserver.cc, tests/api_percentages.cc,tests/api_valuestream.cc: Add some copyright lines missed in changeset [11838]. Mon Feb 16 11:56:46 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * matcher/queryoptimiser.cc: Add another reference from a FIXME to a ticket number. Mon Feb 16 11:47:21 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/omqueryinternal.cc: Update two FIXMEs to reference the appropriate ticket. Mon Feb 16 10:11:39 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * Add support for multi and remote backends with posting sources, from ticket #295: * include/xapian/postingsource.h: Add some clarifying comments about the interface of PostingSources, and add clone(), name(), serialise() and unserialise() methods to PostingSources, for remote and multidatabase support. * include/xapian/query.h: Add internal member to track ownership of external posting sources, and update some comments. * api/postingsource.cc: Add implementations of the new methods, so that all standard PostingSources support remote and multi database searches. * api/omquery.cc: Clone external PostingSources if possible. * api/omqueryinternal.cc: Add serialisation and unserialisation support for posting sources, and support for keeping track of whether an external posting source is owned by the query object or not (ie, whether it needs to be deleted by the destructor). * matcher/externalpostlist.cc,matcher/externalpostlist.h: (Attempt to) clone external posting sources at start, and call the reset() method with the database in use. Delete the posting source afterwards, if the clone was successful. * matcher/queryoptimiser.cc: Pass the database to ExternalPostList. Sadly, this currently requires a const_cast. * tests/: Update tests of PostingSources to take the db parameter on the reset() method, and to test behaviour with multi and remote backends where appropriate. Also, add regression test for segfault in FixedWeightPostingSource::skip_to() when database is empty. * net/remoteserver.cc,common/remoteserver.h: Add the built-in posting sources to the remote server, and add register_posting_source() to the remoteserver, for use in xapian-tcpsrv.cc for user posting sources, analogously to register_weighting_scheme(). Fri Feb 13 20:09:10 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc: Fix segfaults which occur if an empty ValuePostingSource subclass has the skip_to() or check() methods called on it first. * tests/api_db.cc tests/api_valuestream.cc: Move valuemapsource1 to api_valuestream.cc, since that seems like a better place (and api_db.cc is already rather over-large). Add regression test "valuemapsource2" to check the behaviour of ValuePostingSource subclasses on empty lists. Fri Feb 13 19:29:38 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc,include/xapian/postingsource.h: Add ValuePostingSource, which is a base class for implementing posting sources based on reading from a value slot. Convert ValueWeightPostingSource to be a subclass of this, and add ValueMapPostingSource, which is a posting source which uses a map to convert values to weights. * tests/api_db.cc: Add valuemapsource1, testing the ValueMapPostingSource. Fri Feb 13 15:14:53 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/harness/index_utils.cc: Add value 13, containing the first 3 letters of the paragraph - needed for some tests I'm going to commit shortly. Fri Feb 13 14:05:44 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * tests/api_sorting.cc: Change slot number used to get an empty slot from 13 to 100 - we're getting close to using slot 13 now! Sat Feb 07 00:31:12 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_cursor.cc,backends/flint/flint_cursor.cc: Return accidentally changed "XAPIAN_ASSERTIONS_VERBOSE" ifdefs to "XAPIAN_DEBUG_VERBOSE". Sat Feb 07 00:16:43 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * api/omdocument.cc,api/omquery.cc,include/xapian/document.h, include/xapian/query.h: Add methods for serialising documents and queries into strings, and unserialising back from strings. Fixes most of ticket #206 - missing part is that serialising queries containing PostingSources doesn't work. * tests/Makefile.am,tests/api_serialise.cc: Add tests of document and query serialisation. Fri Feb 06 16:06:50 GMT 2009 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Set HAVE_PREAD and HAVE_PWRITE if pread and pwrite are found; turns use of pread and pwrite back on. This improves the speed of a test case of 10,000 (fully cached) searches from 1.84 seconds to 1.78 seconds for me (on ubuntu hardy) - ie, about 3% speed increase. Tue Jan 20 06:22:20 GMT 2009 Olly Betts <olly@survex.com> * tests/api_db.cc: Prefer value.empty() to value == "". Mon Jan 19 05:08:48 GMT 2009 Olly Betts <olly@survex.com> * AUTHORS,queryparser/Makefile.mk: Fix "#line" directives in generated file queryparser/queryparser_internal.cc to give a relative path - previously they had a full path when generated by a VPATH build, and this confused GCC 2.95 and depcomp. Sun Jan 18 23:34:22 GMT 2009 Olly Betts <olly@survex.com> * api/omenquire.cc: Throw UnimplementedError from Enquire::get_mset() if we're asked for a percentage cutoff and to sort primarily by value - this has never been correctly supported and it's better to warn people than give incorrect results. * tests/api_percentages.cc: Add test that this error gets thrown. Tue Jan 13 08:20:36 GMT 2009 Olly Betts <olly@survex.com> * common/omenquireinternal.h: Swap the items vector into the MSet::Internal object rather than copying it to avoid the overhead of the copy. Thu Jan 08 05:14:47 GMT 2009 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.cc,backends/chert/chert_postlist.cc, backends/flint/flint_cursor.cc,backends/flint/flint_postlist.cc, backends/multi/multi_postlist.cc,common/omassert.h,configure.ac, matcher/multimatch.cc: Change preprocessor defines controlling assertions from XAPIAN_DEBUG->XAPIAN_ASSERTIONS and XAPIAN_DEBUG_PARANOID->XAPIAN_ASSERTIONS_PARANOID so their purpose is clearer. Thu Jan 08 02:23:21 GMT 2009 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Wed Jan 07 03:43:07 GMT 2009 Olly Betts <olly@survex.com> * api/matchspy.cc: Fix FP rounding bug (bug#321). * AUTHORS: Add Shane Evans for bug report. Tue Jan 06 22:08:12 GMT 2009 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc: Fix comment typo. Tue Jan 06 21:39:30 GMT 2009 Olly Betts <olly@survex.com> * api/matchspy.cc: Add assertions to try to track down what's causing bug#321. Tue Jan 06 03:20:27 GMT 2009 Olly Betts <olly@survex.com> * matcher/exactphrasepostlist.cc,matcher/phrasepostlist.cc: Fix comment typos. Mon Jan 05 04:42:04 GMT 2009 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 6016" - only change is to comment noting the synced version as the changed code has been deleted in our version. Fri Dec 26 15:05:05 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * include/xapian/database.h: Add Database.close() method, used to close the resources (in particular, the filehandles) held by a database before the destructor is called. Particularly useful for the bindings from other languages. * api/omdatabase.cc,backends/chert/chert_database.cc, backends/chert/chert_database.h,backends/flint/flint_database.cc, backends/flint/flint_database.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/remote/remote-database.cc,common/database.h, common/remote-database.h,net/remoteconnection.cc: Implementation of close() methods. * tests/Makefile.am,tests/api_closedb.cc: New test file, for testing the close method. Could do with being expanded to test many more methods of databases in the closed state, and to test the behaviour after the close of objects created from databases before the database was closed. Fri Dec 26 13:42:14 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: Raise error if database has been closed permanently when cursor_get() is called - a NULL return type isn't checked in various places (for tables which are never closed temporarily), so this is the easiest place to add a check to avoid a segfault without incurring extra overhead. Tue Dec 23 13:42:21 GMT 2008 Olly Betts <olly@survex.com> * languages/compiler/analyser.c: Fix GCC 4.3 warning. Tue Dec 23 06:35:11 GMT 2008 Olly Betts <olly@survex.com> * PLATFORMS: Sync with 1.0 branch. * PLATFORMS: Remove 0.9.x build reports; move 1.0.x build reports to the "old version" section. * INSTALL: Adapt old footnotes from PLATFORMS about HP's aCC and IRIX into new paragraphs here. Tue Dec 23 05:19:52 GMT 2008 Olly Betts <olly@survex.com> * HACKING,configure.ac: Make automake 1.10.2 a hard minimum requirement. Tue Dec 23 04:36:27 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Mon Dec 22 16:02:58 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_databasereplicator.cc, backends/databasereplicator.cc, backends/flint/flint_databasereplicator.cc: Fix some missing bits for debug compilation. Mon Dec 22 13:30:53 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,backends/Makefile.mk,backends/chert/, backends/database.cc,backends/databasereplicator.cc,backends/flint/, common/Makefile.mk,common/database.h,common/databasereplicator.h: Add DatabaseReplicator base class, and subclasses for flint and chert, to hold parts of database replication code which need to be applied to a partially replicated database which may not be a valid database. Move the code for applying changesets and comparing revision numbers into here, and add code for getting uuids which simply returns "" if the database is too broken to have a uuid associated with it. Update the replication code to use DatabaseReplicator subclasses instead of Database classes where necessary, to avoid failing if a partial database is transmitted. Add additional checks to ensure that the UUID of a replicated databse is equal to that sent in the replication protocol message introducing a full database copy; if they differ, a retry of the copy is needed, so ensure that the next message sent is a new copy of the database. Mon Dec 22 11:24:43 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * docs/replication.rst: Warning about a possible way to confuse the replication process. Sat Dec 20 14:00:34 GMT 2008 Olly Betts <olly@survex.com> * tests/api_percentages.cc: Update (C) too. Sat Dec 20 13:51:35 GMT 2008 Olly Betts <olly@survex.com> * matcher/multimatch.cc: If we're using values for sorting and for another purpose, cache the Document::Internal object from getting the value for sorting like we do between other uses. Sat Dec 20 13:09:13 GMT 2008 Olly Betts <olly@survex.com> * tests/api_percentages.cc: Make formatting of MyPostingSource code more self-consistent. Remove a superfluous block scope. Wrap a long comment. Sat Dec 20 12:25:40 GMT 2008 Olly Betts <olly@survex.com> * exception_data.pm: Build a hash mapping a class to a list of its subclasses and export this as %subclasses. Sat Dec 20 05:51:50 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Wrap a long line. Sat Dec 20 05:28:57 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 6016". Fri Dec 19 14:25:10 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 6015". Fri Dec 19 14:07:52 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 6006". Fri Dec 19 13:51:33 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 5919" - code changes are in deleted code, so no actual code changes for us. Fri Dec 19 13:46:22 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 6008". Fri Dec 19 11:15:04 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/api_replicate.cc: Fix comment describing test. Thu Dec 18 23:38:50 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: Tidy up comment wrapping and indenting of code wrapped in try blocks by the previous commit. Thu Dec 18 14:26:30 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_database.cc,backends/chert/chert_table.cc, backends/chert/chert_table.h,backends/flint/flint_database.cc, backends/flint/flint_table.cc,backends/flint/flint_table.h: Improve resilience to unexpected errors during commit. Firstly, fix FlintTable and ChertTable to close the table if an error occurs during commit - this avoids the table being left in an inconsistent state (instead, the table is reopened in a consistent state the next time it is used). Secondly, add a "permanent" close state for tables (indicated by handle being set to -2), which will raise an error if anything tries to do an action which needs the table to be opened, and set the tables to this state if we fail to recover the database after error in commit(). This prevents further actions on the database if we couldn't recover the state (eg, because we're out of disk space), avoiding risk of corruption. Thu Dec 18 01:28:06 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_database.cc,backends/chert/chert_database.h, backends/flint/flint_database.cc,backends/flint/flint_database.h: Improve the error reporting when modifications fail and we cannot set the revision numbers in the table to consistent values, to report the original error too. Wed Dec 17 17:27:24 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: Fix construction of the error message when a base can't be re-read: used to do pointer arithmetic by mistake - now does the correct string concatenation. Wed Dec 17 15:18:10 GMT 2008 Olly Betts <olly@survex.com> * tests/api_valuestream.cc: Skip valuestream3 on chert for now so that snapshots and the tinderbox work again. Wed Dec 17 15:17:07 GMT 2008 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Need omassert.h for STATIC_ASSERT. Wed Dec 17 10:45:45 GMT 2008 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,tests/api_nodb.cc: A NEAR of 2 OR subqueries shouldn't throw AssertionError - instead throw UnimplementedError. Addresses the worst aspect of bug#201, but this should really be implemented. Wed Dec 17 04:44:10 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_database.h: Implement ChertWritableDatabase::open_value_list(). * tests/api_valuestream.cc: Add valueweightsource5 testcase which is a regression test for bug#299 and also exercises the above method. Tue Dec 16 13:30:47 GMT 2008 Olly Betts <olly@survex.com> * common/expandweight.h,expand/expandweight.cc: Overhaul ExpandWeight - now we use the "official" formula and don't return terms which it would give a negative weight to (since that means they are expected to be harmful not helfpul). * tests/api_anydb.cc,tests/api_db.cc: Fix unwarranted assumptions in existing testcases broken by this change. Thu Dec 11 01:02:41 GMT 2008 Olly Betts <olly@survex.com> * examples/delve.cc: Add missing "and" to --help output. Report termfreq and collection freq for each term we're asked about. Mon Dec 08 00:52:51 GMT 2008 Olly Betts <olly@survex.com> * AUTHORS: Add Daniel Andersson for reporting the flintlock permissions issue. Sun Dec 07 22:50:32 GMT 2008 Olly Betts <olly@survex.com> * api/sorter.cc,tests/api_sorting.cc: Fix an empty MultiValueSorter not to SEGV/hang. Fri Dec 05 21:08:08 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,tests/api_backend.cc: Use "flintlock" rather than "chertlock" for the lockfile in chert. The locking is compatible and this avoids the possibility of creating a chert and flint database in the same directory (which will result in one being corrupt since the Btree filenames overlap). Thu Dec 04 09:00:04 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_lock.cc,backends/flint/flint_lock.cc: Create the lockfile with permissions 0666 so that the umask is honoured just like we do for the other files (previously we used 0600). * tests/Makefile.am,tests/api_backend.cc: Add regression test lockfileumask1. Thu Dec 04 08:56:54 GMT 2008 Olly Betts <olly@survex.com> * tests/stemtest.cc: Use str.resize(0) instead of str = "". Thu Dec 04 04:56:55 GMT 2008 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: No need to explicitly initialise a std::string to "". Thu Dec 04 04:55:47 GMT 2008 Olly Betts <olly@survex.com> * tests/api_db.cc: Prefer str.empty() to str == "". Thu Dec 04 01:50:10 GMT 2008 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Make helper function static. Use STATIC_ASSERT to ensure a constant is coprime with 13. Wed Dec 03 22:56:19 GMT 2008 Olly Betts <olly@survex.com> * api/omdatabase.cc: Fix API logging. Wed Dec 03 03:11:45 GMT 2008 Olly Betts <olly@survex.com> * INSTALL: Note that libuuid is required for Xapian 1.1.0 and higher. Wed Dec 03 03:10:59 GMT 2008 Olly Betts <olly@survex.com> * INSTALL: Remove ':' from the end of headings. Mon Dec 01 08:59:04 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Explicitly give the commands to install the required packages for developing on recent Debian or Ubuntu for added convenience. Update the Debian/Ubuntu packages needed for documentation to reflect the change from tetex to texlive. Fix unfinished paragraph about requiring a newline at the end of a source file and note that this is actually undefined behaviour in C++. Sun Nov 30 21:46:59 GMT 2008 Olly Betts <olly@survex.com> * bin/,examples/quest.cc,tests/harness/testsuite.cc: Pull out lists of short options next to list of long options to help make sure they are kept in step. xapian-tcpsrv and xapian-progsrv now accept -w as a short form of --writable, as their long option table had, but their short option string didn't. Sun Nov 30 20:44:48 GMT 2008 Olly Betts <olly@survex.com> * docs/postingsource.rst: Wrap a long line. Note that the remote backend isn't supported yet. Wed Nov 19 11:36:00 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * docs/postingsource.rst: Make it clear that PostingSources must always return documents in increasing document ID order. Wed Nov 19 08:43:04 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/,examples/delve.cc,examples/quest.cc,tests/harness/testsuite.cc: Test return type of gnu_getopt_long for being != -1, rather than == 0, since there are a wide variety of return types other than 0 which may be returned in the successful case (with the builtin implementation in common/getopt.cc, at least). Tue Nov 18 23:22:09 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc: Change unknown "DATABASE" debug type to "DB". Mon Nov 17 13:00:12 GMT 2008 Olly Betts <olly@survex.com> * docs/queryparser.html: Add link to valueranges.html. Fri Nov 14 22:23:10 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update revision we're current to. Fri Nov 14 17:43:15 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_alldocsmodifiedpostlist.cc, backends/chert/chert_modifiedpostlist.cc, backends/flint/flint_modifiedpostlist.cc: Correct comments. Fri Nov 14 17:38:33 GMT 2008 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Update comment about valgrind and C++ STL allocators. Fri Nov 14 17:36:47 GMT 2008 Olly Betts <olly@survex.com> * bin/,examples/delve.cc,examples/quest.cc,tests/harness/testsuite.cc: Don't compare return value of gnu_getopt_long() with EOF - it's documented to simply return -1 or 0, and EOF requires inclusion of stdio.h. Fri Nov 14 17:30:39 GMT 2008 Olly Betts <olly@survex.com> * common/serialise-double.cc: Fix "#elif" with no argument to "#else" - the former surprisingly compiles with most compilers, but GCC 4.4 snapshots reject it. Fri Nov 14 16:54:15 GMT 2008 Olly Betts <olly@survex.com> * common/win32_uuid.cc: A few code clean-ups. Fri Nov 14 13:09:16 GMT 2008 Olly Betts <olly@survex.com> * common/Makefile.mk: Ship common/safeuuid.h to unbreak tarball snapshot building. Sun Nov 09 13:31:19 GMT 2008 Olly Betts <olly@survex.com> * api/omquery.cc,api/omqueryinternal.cc,bin/xapian-tcpsrv.cc, common/omenquireinternal.h,matcher/,net/remoteserver.cc: Prefer "" to <> for including Xapian API headers from within the library. Remove so unused xapian/enquire.h inclusions. Fri Nov 07 12:50:22 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_version.cc,backends/chert/chert_version.h, backends/flint/flint_version.cc,backends/flint/flint_version.h, common/safeuuid.h,common/win32_uuid.cc,common/win32_uuid.h: Add implementation of UUID functions for windows (using the built-in UUID methods in the windows API). Fixes bug #303. Sat Nov 01 01:47:44 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Sync with 1.0.9 release. Fri Oct 31 18:55:03 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/database.h: Document XAPIAN_FLUSH_THRESHOLD (bug#306). Fri Oct 31 18:30:11 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: Fix whitespace inconsistency with omega's configure.ac. Fri Oct 31 10:51:06 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc,include/xapian/postingsource.h: Add a new built-in posting source, FixedWeightSource, which returns every document in the database, but with a fixed weight. In particular, this is useful to add a constant weight to every item in a subquery. Also, in ValueWeightPostingSource, refactor the "last_docid" member to a bool "started", since this is all it's used for. * tests/api_db.cc: Add fixedweightsource1, to test FixedWeightSource. Mon Oct 27 17:05:39 GMT 2008 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Skip test if the timer granularity is too coarse in all cases. Mon Oct 27 08:12:16 GMT 2008 Olly Betts <olly@survex.com> * tests/api_valuestream.cc: Disable valuestream1 for multi backends. Sun Oct 26 00:30:54 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_valuelist.cc: Remove incorrect assertions. Sat Oct 25 06:18:37 GMT 2008 Olly Betts <olly@survex.com> * api/Makefile.mk: Fix typo so we once again ship editdistance.h and maptermlist.h. Thu Oct 23 17:34:13 GMT 2008 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Skip test if the timer granularity is too coarse. Tue Oct 21 05:00:59 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Sync with branches/1.0. Tue Oct 21 02:09:18 GMT 2008 Olly Betts <olly@survex.com> * tests/api_valuestream.cc: Add feature test of ValueIterator::check(). Tue Oct 21 01:36:45 GMT 2008 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h: Prefer string() to "", slot to valno, and not using else after return. Tue Oct 21 01:32:37 GMT 2008 Olly Betts <olly@survex.com> * tests/api_anydb.cc,tests/api_db.cc: Remove '#include <iomanip>' which aren't used. Mon Oct 20 13:30:37 GMT 2008 Olly Betts <olly@survex.com> * api/postingsource.cc,api/valueiterator.cc, include/xapian/postingsource.h,include/xapian/valueiterator.h: Add new API method ValueIterator::check() and use it in PostingIterator. Mon Oct 20 12:28:16 GMT 2008 Olly Betts <olly@survex.com> * tests/api_valuestream.cc: Add test for skip_to() on valuestream iterator. Mon Oct 20 11:19:12 GMT 2008 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/api_valuestream.cc: Add a simple test of valuestream iteration. Mon Oct 20 11:01:26 GMT 2008 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/collate-test,tests/perftest/Makefile.mk: Generate the header listing the collated tests so that we can avoid updating its timestamp when the contents are unchanged, such as in the common case where you modify tests but don't add, remove, or change the conditions on any tests. Mon Oct 20 10:01:56 GMT 2008 Olly Betts <olly@survex.com> * examples/delve.cc: Use valuestream iterator to implement "-V<slot>". Mon Oct 20 02:19:09 GMT 2008 Olly Betts <olly@survex.com> * backends/multi/multi_postlist.cc,backends/multi/multi_postlist.h: Remove unused members of MultiPostList: tname, collfreq_initialised, collfreq, and termweight. Sun Oct 19 14:01:15 GMT 2008 Olly Betts <olly@survex.com> * api/postingsource.cc,include/xapian/postingsource.h: Update ValueWeightPostingSource to use a value stream iterator. Sun Oct 19 13:53:51 GMT 2008 Olly Betts <olly@survex.com> * backends/database.cc: Make use of SlowValueList by default. Sun Oct 19 13:47:10 GMT 2008 Olly Betts <olly@survex.com> * api/documentvaluelist.cc,api/documentvaluelist.h: Fix so that a DocumentValueList starts before the first value. Sun Oct 19 13:09:39 GMT 2008 Olly Betts <olly@survex.com> * api/omdatabase.cc,api/omdocument.cc: Use ValueIterator() instead of ValueIterator(NULL). Sun Oct 19 11:15:15 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Sat Oct 18 01:31:25 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Fri Oct 17 15:54:57 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-compact.cc: Check that pq is not empty before calling top() on it - if it's empty, we only have one database, so we can use the code path which doesn't worry about merging tags. Fixes bug #305. Fri Oct 17 14:22:18 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_valuelist.cc: Fix ChertValueList to start on the right key. Fri Oct 17 14:14:14 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/database.h,include/xapian/document.h: Inline ValueIterator end iterator as ValueIterator() rather than ValueIterator(NULL). * api/valueiterator.cc: Need to call next() on a ValueList to get it to the first position, so do this in ValueIterator's constructor. * api/valueiterator.cc,include/xapian/valueiterator.h: Add ValueIterator::get_docid() method. Fri Oct 17 13:53:26 GMT 2008 Olly Betts <olly@survex.com> * backends/slowvaluelist.cc,backends/slowvaluelist.h: Add missing SlowValueList::get_valueno() method. Fri Oct 17 13:16:01 GMT 2008 Olly Betts <olly@survex.com> * api/omdatabase.cc,include/xapian/database.h: Add Database::valuestream_begin() and Database::valuestream_end() to allow iterating over the values in a given slot for each document in the database (untested so far). Thu Oct 16 13:20:26 GMT 2008 Olly Betts <olly@survex.com> * backends/multi/multi_postlist.cc: Remove unused '#include <list>' and uninformative comment. Thu Oct 16 12:47:00 GMT 2008 Olly Betts <olly@survex.com> * api/omdocument.cc,include/xapian/document.h: Inline Xapian::Document::values_end() into user code. Thu Oct 16 12:26:29 GMT 2008 Olly Betts <olly@survex.com> * api/,backends/chert/chert_valuelist.cc, backends/chert/chert_valuelist.h,common/document.h, common/valuelist.h,include/Makefile.mk,include/xapian/derefwrapper.h, include/xapian/termiterator.h,include/xapian/valueiterator.h: Reimplement ValueIterator to have reference counted internals. One benefit is that Document::Internal no longer need its value_nos map member. Thu Oct 16 11:22:13 GMT 2008 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc: Fix incorrect method name in debug messages. Thu Oct 16 01:34:07 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Note preference for @ rather than \ to introduce doxygen commands. Thu Oct 16 01:21:19 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Expand section on public/protected/private a little. Wed Oct 15 11:34:15 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/database.h: Use std::string() instead of "" as the default value for std::string method parameters. Wed Oct 15 03:20:01 GMT 2008 Olly Betts <olly@survex.com> * backends/Makefile.mk,backends/slowvaluelist.cc, backends/slowvaluelist.h: Support for (slowly) iterating along a value stream for backends which don't support streamed values internally (currently unused). Tue Oct 14 11:43:42 GMT 2008 Olly Betts <olly@survex.com> * api/postingsource.cc,backends/inmemory/inmemory_database.cc, matcher/multimatch.cc: Database::get_document_lazily() now returns NULL if the document isn't found for an inmemory database. * include/xapian/database.h: Document the return value better. Tue Oct 14 03:20:38 GMT 2008 Olly Betts <olly@survex.com> * api/omdatabase.cc,api/postingsource.cc,include/xapian/database.h, matcher/multimatch.cc: Add internal Xapian::Database::get_document_lazily() method and use it in the matcher and Xapian::PostingSource classes to avoid repeating the same code lots of times. Mon Oct 13 23:38:34 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Rename "Miscellaneous Portability Issues" to "C++ Portability Issues", and add a subsection of that for "Miscellaneous Portability Issues" with a new note about needing to terminate the last line of a source file, and a new subsection heading "Header Portability Issues" for all the header-related stuff. Sat Oct 11 12:19:38 GMT 2008 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: If the timer for the first test for bigoaddvalue reports 0.0 seconds, skip the test as the timer doesn't have fine enough granularity for this test to be useful (bug#300). Fri Oct 10 20:40:14 GMT 2008 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Reuse XAPIAN_BIN_PATH to avoid hardcoding more paths, and so this works in a debug build too. Fri Oct 10 17:52:53 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/api_wrdb.cc: Patch to use a different path in win32 build to match the different place that the compiled xapian-check is placed in. Should fix bug #301. Thu Oct 09 05:20:32 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Change -Wstrict-overflow level to 1 - even level 2 causes a few warnings (notably in chert_table.cc) to appear with GCC 4.3 which are unreasonably difficult to avoid. Thu Oct 09 05:04:49 GMT 2008 Olly Betts <olly@survex.com> * HACKING,configure.ac: Raise autoconf requirement to 2.63. Wed Oct 08 14:52:03 GMT 2008 Olly Betts <olly@survex.com> * docs/valueranges.rst: Fix typos in example code. * docs/valueranges.rst,tests/queryparsertest.cc: Drop superfluous empty destructor from ValueRangeProcessor subclass. Wed Oct 08 14:00:04 GMT 2008 Olly Betts <olly@survex.com> * common/valuelist.h: Add missing get_docid() method. Wed Oct 08 12:11:57 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Wed Oct 08 12:00:39 GMT 2008 Olly Betts <olly@survex.com> * backends/Makefile.mk,backends/chert/,backends/database.cc, backends/valuelist.cc,common/Makefile.mk,common/database.h, common/valuelist.h,include/xapian/valueiterator.h: The start of support for iterating along a value stream (currently unused). Tue Oct 07 10:19:48 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: Disable -Wconversion for now - it's not useful for older GCC and is buggy in GCC 4.3. Tue Oct 07 04:53:41 GMT 2008 Olly Betts <olly@survex.com> * matcher/queryoptimiser.h: Query::MatchAll no longer gives match results ranked by increasing document length. * tests/api_db.cc: Extend matchall1 to be a regression test for this. Tue Oct 07 01:04:51 GMT 2008 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Rewrite comment explaining the excess precision fix. Drop unnecessary brackets from return to minimise the diff. Mon Oct 06 07:24:40 GMT 2008 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Change the excess precision fix to pass both values through "volatile double" rather than "float" on platforms where this matters since the former gives better generated code as well as more consistent results with other platforms. Mon Oct 06 07:24:07 GMT 2008 Olly Betts <olly@survex.com> * tests/internaltest.cc: Fix comment typo. Mon Oct 06 02:46:39 GMT 2008 Olly Betts <olly@survex.com> * api/editdistance.cc,api/editdistance.h,api/omdatabase.cc: Pass the largest edit distance we currently care about into the edit distance algorithm so it can terminate early once it knows the edit distance exceeds this. This shortcut is used quite a lot, but there's no measurable speed-up in tests on the real world data I have to hand. Sun Oct 05 12:34:09 GMT 2008 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Adjust percent_factor instead of min_weight so that we don't miss some cases, and make the adjustment actually correspond with the adjustment in omenquire.cc. Wed Oct 01 14:11:49 GMT 2008 Olly Betts <olly@survex.com> * backends/alltermslist.cc,common/alltermslist.h: We don't need an explicit virtual dtor for AllTermsList as it is empty and AllTermsList inherits from TermList which has one. Wed Oct 01 13:21:22 GMT 2008 Olly Betts <olly@survex.com> * common/postlist.h,common/termlist.h: Fix comment typo. Wed Oct 01 00:12:02 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Add a comment for the previous fix. Tue Sep 30 22:41:39 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-compact.cc: Fix bug: if there are both valuestats and metadata, don't overwrite the last metadata value with the first valuestats tag. Tue Sep 30 20:31:34 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Set -Wstrict-overflow to 2 instead of 5, to avoid unreasonable warnings under GCC 4.3. Tue Sep 30 09:45:22 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc: Add missing include, to compile on GCC 4.3.0. Mon Sep 29 13:06:42 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Mon Sep 29 05:04:57 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_version.cc: Add VERSIONFILE_SIZE_LITERAL and use CompileTimeAssert() to ensure it matches the expression in VERSIONFILE_SIZE. It's very rare for the size to change (it's happened once ever) and allows us to build the error string literally at compile time. Sun Sep 28 15:13:11 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_values.cc: Fix bug in lazy update - ensure that the document has loaded its values before deleting the old one (previously, there was an assertion to this effect, but it didn't actually compile, and was also incorrect). * tests/api_wrdb.cc: Check that replacing a document with itself doesn't lose the values - regression test for the bug fixed by this commit. Sun Sep 28 14:04:01 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_values.cc: Fix bug in value updating - modifyvalues1 now passes for chert. Sun Sep 28 12:15:14 GMT 2008 Olly Betts <olly@survex.com> * tests/collate-test: Avoid perl warning when generating files which don't already exist. Sat Sep 27 10:36:09 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/api_wrdb.cc: Add test "modifyvalues1" - regression test for a bug in the streaming values implementation, and also a fairly thorough test of adding and modifying values in databases. Fri Sep 26 19:12:23 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc: Fix warning with gcc 4.2 about possible signed overflow, by using an unsigned type for the total. Fri Sep 26 15:48:39 GMT 2008 Olly Betts <olly@survex.com> * api/omdatabase.cc: Fixed version of: Avoid the relatively expensive edit distance computation for a candidate spelling correction when we can reject the candidate as less good than one we've already seen by looking at the character frequency histograms. In a test on real-world data, this gave an 15% speed-up in queryparsing time with FLAG_SPELLING_CORRECTION set. The cheap check rejects 90% of the words it handles (without the cheap check, the edit distance check rejects 92%). Fri Sep 26 14:40:50 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Check the structure of the value stream chunks. Fri Sep 26 12:21:32 GMT 2008 Olly Betts <olly@survex.com> * tests/api_spelling.cc: Add a regression test for the bug my recently added and reverted spelling optimisation introduced. Fri Sep 26 11:55:18 GMT 2008 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/api_spelling.cc,tests/api_wrdb.cc: Split the spelling tests out into a separate file. Fri Sep 26 11:53:45 GMT 2008 Olly Betts <olly@survex.com> * tests/api_unicode.cc: Capitalise brief file description. Fri Sep 26 11:31:46 GMT 2008 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_document.cc, backends/inmemory/inmemory_document.h: New simpler InMemoryDocument class with slightly reduced memory footprint, based on RemoteDocument class. Fri Sep 26 10:36:02 GMT 2008 Olly Betts <olly@survex.com> * backends/remote/remote-document.cc,backends/remote/remote-document.h, common/document.h: Add new files for RemoteDocument class missed from recent commit. Pass parameter to set_all_values() by non-const reference so we can just efficiently swap() the value maps rather than copying. Fri Sep 26 10:31:21 GMT 2008 Olly Betts <olly@survex.com> * api/omdocument.cc,common/document.h: Remove unused method Xapian::Document::Index::get_all_values(). Fri Sep 26 10:03:17 GMT 2008 Olly Betts <olly@survex.com> * backends/remote/,common/document.h: Replace NetDocument with a new simpler RemoteDocument class which just sets the values and data in its Document::Internal base class. Fri Sep 26 06:21:08 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_document.cc, backends/chert/chert_document.h,common/document.h: Overhaul ChertDocument: Avoid having a reference-counted pointer to the database in the subclass as well as the one now in the base class. Also check the lazy flag before we create a ChertDocument object to avoid a memory allocation in that code path. And make the now-simple constructor inline-able. Thu Sep 25 17:38:37 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * Merge lazyupdate branch: * api/omdocument.cc,backends/chert/chert_database.cc, backends/chert/chert_database.h,backends/chert/chert_values.cc, backends/database.cc,common/database.h,common/document.h: Add a shortcut for calling replace_document() with a document which was just read from the database. If the document terms were never accessed, don't bother updating the termlist, postings and position lists. If the data was never accessed, don't bother updating the data. If the values were never accessed, don't bother updating them. Only fires if we know that the document hasn't been modified (or deleted) since it was read from the database, to ensure that the lazy access of the data doesn't have a visible effect - we enforce this by keeping track of the last document read with open_document(), and only using the shortcut if this is the document passed to replace_document, and that document ID hasn't been passed to replace_document or delete_document since. Add a method Database::Internal::invalidate_doc_object() which is called from the destructor of a document, so that we can reset the pointer to modify_shortcut_document in ChertWritableDatabase when this happens. Document::Internal now keeps a ref-count pointer to the database, to ensure that the database is valid when invalidate_doc_object() is called on it. Thu Sep 25 17:35:07 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * Forward-port from branches/lazyupdate: * tests/api_wrdb.cc: Extend deldoc5 slightly, to check that the DocNotFoundError for a deleted document is raised before the flush, as well as after. Also, test a few more cases which are relevant in the lazyupdate scenario. Thu Sep 25 17:33:21 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * Forward-port from branches/lazyupdate: * backends/chert/chert_document.cc,backends/chert/chert_document.h: Check for existence of a document when opening it by searching for the document length entry, rather than by reading the record for the document. Probably slightly slower if the record is going to be accessed anyway, but much faster if it isn't (though benchmarking of this is needed). Thu Sep 25 14:27:07 GMT 2008 Olly Betts <olly@survex.com> * Merge valuestreams branch: * backends/chert/,bin/xapian-check.cc,bin/xapian-compact.cc, docs/admin_notes.rst: Change the chert backend to store values in a chunked stream for each value slot to make access more efficient. The chunked streams and value stats are now stored in the postlist table while the list of used value slots for each document is stored in the termlist table, so we no longer need a special table for values. * bin/xapian-check.cc,bin/xapian-check-flint.cc: Don't stop checking a table after an error in certain cases - instead increment the error counter and try to continue checking from the next item. Thu Sep 25 13:50:15 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/queryoptimiser.cc,tests/api_anydb.cc: Apply fix for platforms with excess precision - this fixes a potential segfault on x86 architecture when doing OP_ELITE_SET queries. Also, add a regression test. Thu Sep 25 03:38:24 GMT 2008 Olly Betts <olly@survex.com> * examples/copydatabase.cc: Also copy user metadata. Wed Sep 24 17:27:33 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc: Back out, for now, Olly's change from yesterday to avoid some edit distance checks; counting the frequency of letters doesn't take into account the number of substitutions, so spelling corrections which involve substitutions are missed. Wed Sep 24 12:48:39 GMT 2008 Olly Betts <olly@survex.com> * unicode/tclUniData.cc: Update to Unicode 5.1 (bug#265). * tests/api_unicode.cc: Add tests for some of the changes in Unicode 5.1. Tue Sep 23 14:58:54 GMT 2008 Olly Betts <olly@survex.com> * api/omdatabase.cc: Avoid the relatively expensive edit distance computation for a candidate spelling correction when we can reject the candidate as less good than one we've already seen by looking at the character frequency histograms. Mon Sep 22 09:21:55 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * Forward-port from branches/lazyupdate: * backends/chert/chert_postlist.cc: Fix return value of jump_to() to be true only if the exact item specified was found in the doclen list (previously also returned true if an item following the item being searched for was found in the chunk). Note that there is no test for this fix included in this forward-port, but an implicit test is included in the lazyupdate branch, and will be here when that branch is merged to trunk. Mon Sep 22 07:58:06 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * HACKING: Mention that setting XAPIAN_DEBUG_LOG to '-' has a special effect. Mon Sep 22 06:27:23 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * docs/remote_protocol.html: Update documentation of remote protocol to mention UUID being passed in REPLY_GREETING and REPLY_UPDATE (added in revision [11369]). Thu Sep 18 11:52:05 GMT 2008 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Fix cursordelbug1 to hopefully work on Microsoft Windows. Thu Sep 18 11:50:04 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.cc,backends/chert/chert_table.cc, backends/chert/chert_table.h: More assertions and improved debug logging coverage. Use rare() to mark a couple of checks which almost always have the same result. Make an assertion about the level of a Btree block into an always-compiled-in check which throws an exception, as this isn't a sanity check of internal state. Thu Sep 18 05:26:27 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_postlist.cc,backends/flint/flint_postlist.cc: Fix comment describing format of postlist chunk (collection freq was missing). Wed Sep 17 12:38:51 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Fix typos in XAPIAN_DEBUG_FLAGS documentation. Sat Sep 13 11:14:02 GMT 2008 Olly Betts <olly@survex.com> * HACKING,configure.ac: Upgrade libtool requirement to 2.2.6a. Fri Sep 05 11:40:55 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/api_replicate.cc: Test that the UUID of a replica is equal to the UUID of the original database. Fri Sep 05 09:38:35 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * include/xapian/database.h,api/omdatabase.cc: Add Database.get_uuid() which returns a unique identifier for the database. * tests/api_anydb.cc: Add test uuid1 to test basic get_uuid() behaviour. * common/database.h: Extend documentation comment for get_uuid() a bit. * backends/remote/remote-database.cc,common/remote-database.h, net/remoteserver.cc: Add support for get_uuid() to remote databases. Thu Sep 04 11:37:54 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am: Remove accidentally committed reference to apitest_sortconsist.txt, which only exists in my local checkout and in a patch in ticket #216. Thu Sep 04 04:39:32 GMT 2008 Olly Betts <olly@survex.com> * HACKING,configure.ac: We now hard require autoconf 2.62, so it's safe to use AC_TYPE_SSIZE_T. (bug#135) Thu Sep 04 04:26:59 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: Set version to 1.1.0. Thu Sep 04 03:32:06 GMT 2008 Olly Betts <olly@survex.com> * AUTHORS,NEWS,configure.ac: Sync with 1.0.8. Wed Sep 03 15:23:54 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am,tests/api_percentages.cc: Add new test file for tests related to percentage weights (there are a few around which could be moved here, and there's another one in the patch in ticket #216). Add test in this file which uses a PostingSource with carefully constructed weights to check the rounding behaviour for percentage cutoffs. Wed Sep 03 07:58:29 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: Adjust min_weight values calculated for percent_cutoff in the same manner as in omenquire.cc, to allow for excess precision in the same way. Wed Sep 03 07:15:49 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_version.cc: Display numerical value of correct size of chert version file in error message if size is incorrect, rather than a complicated expression giving the correct size. Tue Sep 02 10:41:46 GMT 2008 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Remove FIXME for pushing check_at_least handling into the remote backend, which was done for 1.0.2. Tue Sep 02 06:48:50 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-inspect.cc: Show the help message on start-up. Correct the alias for next from ' ' to ''. Avoid reading outside of input string when it is empty. Bug#286. Tue Sep 02 04:19:34 GMT 2008 Olly Betts <olly@survex.com> * tests/stemtest.cc: Handle the new supplemental stemming data. Tue Sep 02 03:05:10 GMT 2008 Olly Betts <olly@survex.com> * tests/stemtest.cc: Update for new xapian-data directory structure. Tue Aug 26 09:26:49 GMT 2008 Olly Betts <olly@survex.com> * common/debuglog.h: Handle logging from ctor, dtor, and void-returning functions/methods with a separate class (DebugLogFuncVoid). Fix outdent handling for non-logged message types. Fix detection of active exceptions to check if the exception was already active when we entered the current function. This probably means that "RETURN_VOID" is no longer needed, so comment it out. Tue Aug 26 07:34:16 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Document explicitly that XAPIAN_DEBUG_FLAGS=- gives you all debug messages. Tue Aug 26 00:57:32 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Fix reversed check for whether a table is in a flint or chert database. Mon Aug 25 12:03:36 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_termlisttable.cc: Remove FIXME about dropping something if we make an incompatible database version bump - that's not going to happen now for flint. Avoid packing doclen for an empty termlist. Add explicit cast for char for prev_term.size(). Sun Aug 24 11:42:21 GMT 2008 Olly Betts <olly@survex.com> * common/Makefile.mk: Fix build with --enable-log=profile. Sun Aug 24 12:41:22 BST 2008 Olly Betts <olly@survex.com> * HACKING: Add a bit of discussion of --enable-log=profile. Tweak formatting in one place and wording in another. Sat Aug 23 00:48:40 GMT 2008 Olly Betts <olly@survex.com> * tests/collate-test: Only update files which have changed to avoid rebuilding all apitest's source files when only one has changed. Fri Aug 22 01:02:57 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 5571" (no code changes in our modified version). Fri Aug 22 01:00:15 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Fix comment typo. Fri Aug 22 00:55:26 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 5564". Fri Aug 22 00:44:14 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Merge upstream "Check-in Number: 5488" and "5563". * queryparser/queryparser.lt: Fix compilation for previous merge (I fixed it in the generated file before!) Fri Aug 22 00:07:22 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 5472". Thu Aug 21 04:15:41 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: Remove code which checks for --enable-debug and --enable-debug-verbose (deprecated since 1.0.0) and gives an error pointing to the replacements. * docs/deprecation.rst: Document the replacements here. Thu Aug 21 00:57:52 GMT 2008 Olly Betts <olly@survex.com> * common/omdebug.h: Remove DEBUGLINE completely. * backends/flint/flint_table.cc,backends/inmemory/inmemory_database.cc, queryparser/queryparser.lt: Replace remaining DEBUGLINE uses. * backends/inmemory/inmemory_database.cc: Don't need <list>. Thu Aug 21 00:37:08 GMT 2008 Olly Betts <olly@survex.com> * common/debuglog.h,common/omdebug.h: Define LOGLINE, etc to no-op versions when XAPIAN_DEBUG_VERBOSE isn't defined. * api/,backends/chert/,backends/flint/, backends/inmemory/inmemory_database.cc, backends/multi/multi_postlist.cc,backends/multi/multi_termlist.cc, expand/esetinternal.cc,expand/expandweight.cc,matcher/, net/remoteconnection.cc: Use LOGLINE or LOGVALUE instead of DEBUGLINE. Wed Aug 20 13:33:56 GMT 2008 Olly Betts <olly@survex.com> * common/debuglog.h: Don't need omtime.h. Wed Aug 20 06:11:31 GMT 2008 Olly Betts <olly@survex.com> * examples/quest.cc: Fix to catch QueryParserError instead of const char * which Xapian < 1.0.0 threw instead. Wed Aug 20 05:50:26 GMT 2008 Olly Betts <olly@survex.com> * docs/valueranges.rst: Expand on some sections. Wed Aug 20 04:43:07 GMT 2008 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Clarify in comments that this affected 1.0.7. Wed Aug 20 04:33:22 GMT 2008 Olly Betts <olly@survex.com> * api/omenquire.cc: Simplify RSet::remove_document() and RSet::contains() a little. Fix output of RSet::get_description(). * tests/api_nodb.cc: Add regression test rset4 for RSet::get_description() fix. Wed Aug 20 02:45:04 GMT 2008 Olly Betts <olly@survex.com> * common/omdebug.h: Remove DebugMsg() macro which is no longer used. Tue Aug 12 05:03:34 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc,backends/flint/flint_table.h: Backport elimination of find_tag() to flint. Tue Aug 12 04:10:41 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc,backends/chert/chert_table.h: ChertTable::find_tag() is only used by ChertTable::get_exact_entry() so remove the former folding its code into the latter. Mon Aug 11 03:06:10 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_spelling.h,backends/chert/chert_synonym.h: Remove random closing bracket from a couple of comments. Mon Aug 11 03:03:19 GMT 2008 Olly Betts <olly@survex.com> * api/postingsource.cc: Always initialise max_value in the constructor body for consistency. Sat Aug 09 09:56:06 GMT 2008 Olly Betts <olly@survex.com> * tests/api_valuestats.cc: Add valuestats4 as a regression test for the previous fix. Sat Aug 09 06:10:49 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_database.h: Write cached valuestats changes to disk when we automatically flush changes due to the number of documents indexed. Wed Aug 06 11:27:29 GMT 2008 Olly Betts <olly@survex.com> * api/omdocument.cc,backends/chert/chert_document.cc, backends/chert/chert_document.h,backends/flint/flint_document.cc, backends/flint/flint_document.h, backends/inmemory/inmemory_document.cc, backends/inmemory/inmemory_document.h, backends/remote/net_document.cc,backends/remote/net_document.h, common/document.h: Change do_get_all_values() to pass a reference to the std::map to return the result in rather than returning the std::map. Wed Aug 06 07:39:08 GMT 2008 Olly Betts <olly@survex.com> * docs/replication.rst: Adding missing (C) for new section addded recently. Wed Aug 06 06:27:59 GMT 2008 Olly Betts <olly@survex.com> * backends/alltermslist.cc,backends/chert/,backends/flint/, backends/inmemory/inmemory_alltermslist.cc, backends/inmemory/inmemory_alltermslist.h, backends/multi/multi_alltermslist.cc,common/alltermslist.h, common/multialltermslist.h: Add a default "not implemented" implementation of AllTermsList::get_approx_size() and remove the "not implemented" versions in all the subclasses. Remove empty destructors for ChertSpellingTermList, and FlintSpellingTermList. * api/maptermlist.h,api/omdocument.cc: MapTermList::get_approx_size() should never be used, so make it "Assert(false); return 0;". Wed Aug 06 02:24:48 GMT 2008 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/multi/multi_alltermslist.cc, common/multialltermslist.h: Reimplement MultiAllTermsList to use a heap rather than linearly scanning - operations during iteration should now be O(log(n)) rather than O(n) where n is the number of databases. Tue Aug 05 23:50:19 GMT 2008 Olly Betts <olly@survex.com> * tests/Makefile.am: Fix "make clean" to remove cached databases again and "make check" to remove cached databases like the previous change aimed to do. Tue Aug 05 23:48:52 GMT 2008 Olly Betts <olly@survex.com> * tests/api_db.cc: Use TEST_EQUAL(a, b) rather than TEST(a == b). Tue Aug 05 12:44:58 GMT 2008 Olly Betts <olly@survex.com> * common/alltermslist.h: Fix parameter name in doxygen comment. Tue Aug 05 06:28:15 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/flint/flint_database.cc: Don't generate a changeset file for the first revision of a database (since we might as well just send the whole database) - addresses part of bug#278. Clean up exceptions to pass errno in the errno parameter rather than incorporating strerror(errno) in the message parameter. Tue Aug 05 06:00:11 GMT 2008 Olly Betts <olly@survex.com> * tests/api_replicate.cc: Remove "sleep(1);" from testcase replicate1 which is no longer required now we generate proper UUIDs. Tue Aug 05 03:07:11 GMT 2008 Olly Betts <olly@survex.com> * api/replication.cc,include/xapian/replication.h: Remove DatabaseReplica::get_parameter() and set_parameter() as they're no longer used for storing the database UUID, and Richard said the thinking behind wanting them was "flawed". Tue Aug 05 02:03:24 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_version.cc: Comment tweak. Tue Aug 05 02:02:29 GMT 2008 Olly Betts <olly@survex.com> * api/replication.cc,backends/flint/flint_database.cc,common/utils.cc, common/utils.h: Use the UUID stored by the backend rather than storing it ourselves as a database replica parameter. Mon Aug 04 14:58:21 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_version.cc, backends/flint/flint_version.h: Generate a UUID when a new flint database is created, and lazily generate one for existing flint databases which don't have one. Store the UUID in a new "uuid" file in the database directory to avoid having to change the flint database format incompatibly. Mon Aug 04 14:13:53 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_version.cc, backends/chert/chert_version.h,configure.ac: Generate a UUID when a chert database is created, and store it in the version file. Mon Aug 04 12:16:54 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_version.cc: Make exception messages more consistent. Mon Aug 04 12:15:14 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_version.cc, backends/chert/chert_version.h: Remove the code to handle upgrading flint versions >= 200704230 and < 200709120 as it's just dead code for chert. Sun Aug 03 14:34:31 GMT 2008 Olly Betts <olly@survex.com> * api/replication.cc: Track the replica numbers using an int rather than two strings. Eliminate two racey uses of file_exists(). Sun Aug 03 11:37:06 GMT 2008 Olly Betts <olly@survex.com> * docs/replication.rst: Add section on backend support. Make FIXME a comment. Fix erroneous reference to '"-h" parameter'. Sun Aug 03 11:28:27 GMT 2008 Olly Betts <olly@survex.com> * api/replication.cc,tests/harness/testrunner.cc: Make replication work for the chert backend. Sat Aug 02 16:47:04 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * include/xapian/enquire.h: Correct documentation comment (iterator doesn't return terms, it returns mset items). Sat Aug 02 14:16:24 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * docs/admin_notes.rst: Fix a typo. Sat Aug 02 14:06:37 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.cc,backends/flint/flint_cursor.cc: Fix comment typo in previous commit. Sat Aug 02 13:04:05 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_cursor.cc,backends/flint/flint_cursor.cc: Fix FlintCursor::del() and ChertCursor::del() to leave the cursor on the next key when iterating over the unflushed revision (previously the cursor would end up an extra key along in this case). (Bug#287) * tests/api_wrdb.cc: Add regression test cursordelbug1. Sat Aug 02 05:09:38 GMT 2008 Olly Betts <olly@survex.com> * api/replication.cc,common/utils.cc,common/utils.h: Change removedir() to just return if passed a non-existent path and document this. Make use of this to avoid having to call dir_exists() before remove_dir(). Sat Aug 02 04:43:45 GMT 2008 Olly Betts <olly@survex.com> * tests/api_db.cc,tests/apitest.cc: Make use of new BackendManager::get_database_path() to eliminate knowledge of BackendManagerFlint's implementation from test stubdb1. Put the stubdb files in a .stub subdirectory and don't remove them after each test to aid debugging failing tests. Also now run stubdb tests under chert and multi backends. Add more stubdb tests for the recently added features. Sat Aug 02 04:39:08 GMT 2008 Olly Betts <olly@survex.com> * tests/harness/: Adjust the BackendManager interface to reduce code duplication in subclasses - for backends where the concept of a path to the database makes sense, we now support a get_database_path() method and use this in the default implementation of get_database() in the base class. Sat Aug 02 04:36:48 GMT 2008 Olly Betts <olly@survex.com> * tests/apitest.h: Use std::string() in preference to "". Sat Aug 02 04:35:13 GMT 2008 Olly Betts <olly@survex.com> * backends/dbfactory.cc: Fix handling of new "inmemory" type in stub databases. Sat Aug 02 01:49:35 GMT 2008 Olly Betts <olly@survex.com> * tests/Makefile.am: Remove the cached test databases in before running the testsuite by making check-local do the same as clean-local. Sat Aug 02 01:42:36 GMT 2008 Olly Betts <olly@survex.com> * backends/dbfactory.cc: Add support for "inmemory" to stub database files (it's useful now they can be writable). Don't just ignore lines with no spaces in in stub database files. Thu Jul 31 13:13:30 GMT 2008 Olly Betts <olly@survex.com> * backends/Makefile.mk,backends/database.cc,backends/dbfactory.cc, common/database.h,include/xapian/dbfactory.h: Add new Auto::open_stub() overload which opens a stub database file containing a single entry as a WritableDatabase. Move the non-remote database factory API functions from backends/database.cc to new file backends/dbfactory.cc, leaving the former solely for methods of Database::Internal. A stub database file is now allowed to contain no database entries, which results in an empty Database object (this avoids user code having to special case to handle "0 or more" databases). * tests/api_nodb.cc: Tidy up #include directives. Enhance nosuchdb1 to check that we get a reasonable error message - previously we got "Couldn't detect type of database". * include/xapian/dbfactory.h: Improve doxygen comment for InMemory::open(). Thu Jul 31 09:36:41 GMT 2008 Olly Betts <olly@survex.com> * tests/api_replicate.cc: This is a new file, so use <cstdlib> rather than <stdlib.h>. Wed Jul 30 23:10:19 GMT 2008 Olly Betts <olly@survex.com> * backends/database.cc: Bad lines in a stub file were being ignored after we'd seen a good entry. * tests/api_db.cc: Add regression test. Wed Jul 30 14:42:04 GMT 2008 Olly Betts <olly@survex.com> * generate-exceptions.in: Use std::string() in preference to "". Tue Jul 29 23:55:21 GMT 2008 Olly Betts <olly@survex.com> * AUTHORS: Add Henrik Brix Andersen. Tue Jul 22 06:01:44 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc,backends/flint/flint_table.h: Eliminate other_base_letter member of FlintTable - its value can always be easily determined from base_letter. Mon Jul 21 12:33:54 GMT 2008 Olly Betts <olly@survex.com> * xapian-config.in: Add bug report URL to xapian-config --help output. Mon Jul 21 11:55:18 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: Put the bug report URL as the third parameter to AC_INIT. Add proper m4 quoting in a few places (nowhere that should actually change behaviour). Mon Jul 21 01:03:12 GMT 2008 Olly Betts <olly@survex.com> * tests/apitest.cc,tests/harness/testsuite.cc, tests/harness/testsuite.h: Report subtotals per backend, rather than per testgroup per backend to make the output much clearer to scan. Mon Jul 21 00:36:07 GMT 2008 Olly Betts <olly@survex.com> * tests/apitest.cc: Use startswith(). Sun Jul 20 14:16:18 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/harness/testsuite.h: Add missing "iomanip" header needed for setprecision calls. Sun Jul 20 11:35:35 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/enquire.h,include/xapian/matchspy.h, include/xapian/valueiterator.h: Remove explicit empty non-virtual destructors since the compiler will create them by default anyway. Sun Jul 20 10:18:52 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Sun Jul 20 10:09:36 GMT 2008 Olly Betts <olly@survex.com> * Forward-port change from branches/1.0: * api/omenquire.cc: Fix percentage calculation to cope with excess precision on x86. Fix method name for convert_to_percent_internal in debug logging. Sun Jul 20 09:13:32 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/,bin/xapian-check-flint.cc,bin/xapian-check-flint.h, bin/xapian-compact.cc: Store the tablename in FlintTable as a const char * - it's a constant string and std::string adds a needless space overhead. Sun Jul 20 09:10:49 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Follow-on fix for last change - we need to use strcmp() to compare C strings. Sun Jul 20 07:09:38 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/,bin/xapian-check.cc: Store the tablename in ChertTable as a const char * - it's a constant string and std::string adds a needless space overhead. Sat Jul 19 14:19:58 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_values.cc, backends/chert/chert_values.h: Fix WritableDatabase::add_document() and replace_document() not to be O(n*n) in the number of values in the new document. * backends/flint/flint_database.cc,backends/flint/flint_values.cc, backends/flint/flint_values.h: Backport fix to flint. * tests/api_wrdb.cc: Add testcase bigoaddvalue to make sure we don't regress to O(n*n) (or worse!) Fri Jul 18 13:24:43 GMT 2008 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Rewrite lazytablebug1 testcase to avoid having to run xapian-compact. Fri Jul 18 11:59:42 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc: Fix handling of a table created lazily after the database has had commits, and which is then cursored while still in sequential mode. * backends/flint/flint_table.cc: Backport fix to flint. * tests/api_wrdb.cc: Add testcase lazytablebug1. Fri Jul 18 02:52:55 GMT 2008 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Enable test crashrecovery1 for chert. Finish off unfinished comment. Fri Jul 18 02:38:22 GMT 2008 Olly Betts <olly@survex.com> * tests/api_wrdb.cc,tests/harness/testrunner.cc, tests/harness/testrunner.h: Add "synonyms" test backend property and use it to decide where to run synonym tests so they get run on chert too. Move virtual TestRunner destructor out of the header. Don't skip all the spelling tests on chert. Fri Jul 18 00:56:17 GMT 2008 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Eliminate literal top-bit-set characters. Fri Jul 18 00:40:06 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_values.cc: Kill superfluous else tokens. Thu Jul 17 13:00:55 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_table.cc,backends/chert/chert_table.h: Eliminate other_base_letter member of ChertTable - its value can always be easily determined from base_letter. Thu Jul 17 12:24:19 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc: Add space after catch for consistency. Thu Jul 17 11:51:57 GMT 2008 Olly Betts <olly@survex.com> * examples/copydatabase.cc: Use C++ forms of C headers. Only treat '\' as a directory separator on platforms where it is. Update counter every 13 counting up to the end so that the digits all "rotate" and the counter ends up on the exact total. Wed Jul 16 10:10:32 GMT 2008 Olly Betts <olly@survex.com> * tests/perftest/perftest.cc: Use uname() or gethostname() to get the hostname on Unix, rather than piping output from uname -n. Use startswith() where appropriate. Prefer resize() and erase() to assigning a string a substring of itself. Make internal functions static. Prefer string() to "". Fold sed | sed and grep | sed into single sed invocations. Wed Jul 16 05:16:00 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Merge upstream "Check-in Number: 5404" and "5405". Wed Jul 16 04:48:45 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Sync with 1.0.7 release. Start to clean up ChangeLog entry pile. Mon Jul 14 05:13:55 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Fix a typo, remove "\n" from the end of ErrorMsg() calls since it will get stripped anyway, and escape literal '%' in ErrorMsg() third arguments. Sun Jul 13 13:04:58 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Update note of upstream version we're synced against. Sun Jul 13 12:56:58 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Tweak previous merged patch so that we still get #line directives for code in %include directives. Sun Jul 13 12:54:03 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 5335". Reduces the size of the stripped object file for queryparser_internal.cc by about 0.5%. Sun Jul 13 12:32:06 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Fix typo in comment. Sun Jul 13 12:30:38 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 5334". No change to the generated code in our case. Sun Jul 13 12:24:13 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: Add missing hard requirement for libtool 2.2.4. Sun Jul 13 12:09:40 GMT 2008 Olly Betts <olly@survex.com> * configure.ac,HACKING: Hard require autoconf 2.62 and automake 1.10.1. Move all information about particular autoconf and automake versions to HACKING. * configure.ac: Use LT_INIT in preference to AC_PROG_LIBTOOL. On Linux and k*bsd-gnu, override libtool's link_all_deplibs_CXX to "no". On Linux, override libtool's sys_lib_dlsearch_path_spec to a list generated in a more reliable way which includes *all* the default directories. Remove workaround which sets docdir for autoconf < 2.60 since we now require 2.62. * xapian-core.spec.in: We no longer need to run autoreconf to work around libtool's incomplete sys_lib_dlsearch_path_spec or to pick up distro-specific patches for link_all_deplibs. Sun Jul 13 11:33:33 GMT 2008 Olly Betts <olly@survex.com> * Makefile.am: No need to explicitly list m4/*.m4 in EXTRA_DIST as automake will automatically ship any such files which are needed. Sun Jul 13 09:15:34 GMT 2008 Olly Betts <olly@survex.com> * api/omenquire.cc: Back out the rounding of percentages change (at least for now), as it makes percentage cut-offs work inconsistently. * tests/api_anydb.cc: Add test pctcutoff3 which demonstrates the issue. Sun Jul 13 07:01:02 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Fix documentation comment for MSet::get_rank() - the 'document judged "most relevant" will have rank of 0' only when ordering primarily by relevance. Sun Jul 13 06:17:13 GMT 2008 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Tweak layout (makes the backport diff much smaller and clearer, and reduces indentation by a level for a chunk of code). Fri Jul 11 13:44:56 GMT 2008 Olly Betts <olly@survex.com> * tests/harness/testutils.cc: Fix typos in recent improved output. Fri Jul 11 13:04:41 GMT 2008 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Collect up all the scaling factors we apply to the estimated number of matches and apply them in one go to avoid rounding the result more than once. Fri Jul 11 12:21:11 GMT 2008 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Add testcase for percent cutoff plus collapsing which most likely would have failed before Richard's recent fix for the lower bound with collapsing and a matchdecider. Fri Jul 11 05:30:02 GMT 2008 Olly Betts <olly@survex.com> * backends/database.cc: Add support for XAPIAN_PREFER_CHERT environmental variable. Thu Jul 10 22:21:46 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc: Fix skip_to() so that it doesn't advance if it's already in a suitable position. * include/xapian/postingsource.h: Specify that skip_to() should stay in the same position if the current position is equal to the docid argument of skip_to(). * tests/api_db.cc: Add valueweightsource3 to check that skip_to() stays in the same position in this case. Thu Jul 10 13:14:14 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc: Performance improvement - open documents lazily (using the internal interface) so that we don't access the record table for each one - we now just access the value table. Leads to considerable speed up for me (of the order of 5 times faster). Thu Jul 10 09:49:39 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Overhaul the sections on building from SVN. Remove ':' from the end of headings. Wed Jul 09 10:43:01 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: The workaround to avoid probe code for F77, GCJ, and RC being added to configure is no longer required now that we're using libtool 2.2 so remove it. Wed Jul 09 09:40:43 GMT 2008 Olly Betts <olly@survex.com> * Makefile.am,acinclude.m4,autoconf/dir_contents, autoconf/rjb_find_stlport.m4,autoconf/type_socklen_t.m4, m4/dir_contents,m4/rjb_find_stlport.m4,m4/type_socklen_t.m4: Move the m4 macros which we ship for building configure from into the m4 subdirectory and remove acinclude.m4 - now aclocal will pull in the required macros automatically. Wed Jul 09 09:14:31 GMT 2008 Olly Betts <olly@survex.com> * Makefile.am: Update for m4 -> m4-macros change. Wed Jul 09 09:11:45 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/externalpostlist.cc: Fix segfault if get_maxweight() is called when the postlist has reached end. This happens if the end is reached during decay of an AND_MAYBE postlist to an AND. * common/postlist.h,matcher/: Add comments documenting that recalc_maxweight() may be called after the postlist has reached the end, and noting why this is safe in various cases. Wed Jul 09 09:02:08 GMT 2008 Olly Betts <olly@survex.com> * m4-macros/dir_contents,m4-macros/xapian.m4,m4/dir_contents, m4/xapian.m4: Move xapain.m4 to a new subdirectory (m4-macros) since libtoolize now installs m4 files into subdirectory m4 which we don't really want picking up by applications building against an uninstalled xapian-core. Wed Jul 09 08:03:11 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * Makefile.am,m4/dir_contents,xapian.m4: Move xapian.m4 back into the m4 subdirectory. For uninstalled builds, we need xapian.m4 to be in a directory without any other .m4 files, or we'll be likely to pull in the wrong versions of other macros. The top level directory contains aclocal.m4 and acinclude.m4, so isn't suitable for this. Also, the xapian-bindings and omega compilations hadn't been updated with the new xapian.m4 location, so this fixes the build for them. Wed Jul 09 07:17:38 GMT 2008 Olly Betts <olly@survex.com> * Makefile.am,configure.ac: Use AC_CONFIG_MACRO_DIR and ACLOCAL_AMFLAGS as libtoolize 2.2.4 recommends. Wed Jul 09 06:37:21 GMT 2008 Olly Betts <olly@survex.com> * acinclude.m4: Remove reference to definedir.m4. Wed Jul 09 05:32:24 GMT 2008 Olly Betts <olly@survex.com> * Makefile.am,m4/dir_contents,m4/xapian.m4,xapian.m4: Move xapian.m4 to the top level rather than devoting a whole directory to a single file. Wed Jul 09 05:23:02 GMT 2008 Olly Betts <olly@survex.com> * Makefile.am,autoconf/definedir.m4,configure.ac: AC_DEFINE_DIR() hasn't been used for years so remove it. Wed Jul 09 02:32:38 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc,include/xapian/postingsource.h: Add an alternative constructor for ValueWeightPostingSource which allows the upper bound on the weights stored to be specified manually. This is useful for database formats like flint which don't have an upper bound available. It could conceivably also be useful if you know that a query will only be accessing a subset of documents for which you know a more precise upper bound than the database-wide upper bound. Wed Jul 09 01:59:56 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/externalpostlist.cc: Add DEBUGCALL macros to enable easier debugging. Mon Jul 07 12:40:20 GMT 2008 Olly Betts <olly@survex.com> * HACKING,NEWS: Update to reflect the autotools now being in the SVN tree, and the versions we are now bootstrapping with. * NEWS: Update from ChangeLog. Sun Jul 06 22:59:04 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc,matcher/remotesubmatch.cc, matcher/remotesubmatch.h: Apply patch from ticker #279 to improve performance of matches with multiple remote databases. This adds a special case for the top match being from a remote database, to avoid accessing the termlist to calculate the percentage scaling factor. Sun Jul 06 22:55:17 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omenquire.cc: When calculating percentages, round to the nearest integer, rather than rounding down. There was a FIXME about this, but no explanation of why it hadn't already been done, and I can see no bad side effects so far. The most obvious positive effect is that queries which should get precisely 100% will no longer be assigned 99% due to rounding errors. Sun Jul 06 22:48:30 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/api_anydb.cc: Add new test, topercent2, checking the percentage values returned by standard searches, both against known current values for some of the hits, and against a search with a "local" backend. Also, in rsetmultidb1, test the return value of mset_range_is_same_weights(). * tests/api_db.cc: In rsetmultidb2, test the return value of mset_range_is_same_weights(). Sun Jul 06 22:46:23 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/harness/testutils.cc,tests/harness/testutils.h: Add function to test if two mset ranges have the same percentages, and slightly improve the messages on failure of the mset range comparison tests. Fri Jul 04 23:31:14 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * NEWS: Swap two words to make a sentence make sense. Fri Jul 04 11:40:13 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: Fix calculation of lower bound when collapsing and match deciders are used. (Only tended to manifest when the collapsing threw away a lot of documents, and when also using a custom sort order, but I don't think that was required.) Also, perform the adjustments due to percentage cutoffs after the other adjustments, to avoid the lower bound setting for collapsing overriding the lower bound from the percentage cutoff. * tests/harness/index_utils.cc: Add a value to value slot 12, which only has 5 different values, for testing this bug. * tests/api_db.cc: Add regression test for calculation of the lower bound. Fri Jul 04 10:24:16 GMT 2008 Olly Betts <olly@survex.com> * Makefile.am,bin/Makefile.mk,examples/Makefile.mk: Remove extra_cleandirs as automake 1.10.1 fixes the bug which it was working around. Fri Jul 04 04:36:59 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Synchronise tiny change from branches/1.0 which isn't on trunk for some reason. Fri Jul 04 04:35:19 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Bootstrap with newer versions of the autotools: + autoconf 2.61 -> 2.62: faster + automake 1.10 -> 1.10.1: "make clean" cleans all .libs directories. + libtool 1.5.24 -> 2.2.4: many improvements; faster. Thu Jul 03 03:17:42 GMT 2008 Olly Betts <olly@survex.com> * docs/glossary.rst,docs/intro_ir.html: Improve intro_ir a bit, and link to the definition of RSet in the glossary. Thu Jul 03 01:30:56 GMT 2008 Olly Betts <olly@survex.com> * examples/quest.cc: Output get_description() of the parsed query. Sun Jun 29 18:56:06 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: Rename percent_factor variable, used here as part of the percent_cutoff mechanism, to make its purpose clearer, and to avoid confusing it with percent_scale in this file, and percent_factor used elsewhere in matcher/. Sat Jun 28 21:51:42 GMT 2008 Olly Betts <olly@survex.com> * docs/doxygen_api_conf.in,docs/doxygen_full_conf.in: Disable "JAVADOC_AUTOBRIEF" since we always try to write a brief description explicitly, and it causes problems in some cases. Sat Jun 28 21:32:21 GMT 2008 Olly Betts <olly@survex.com> * common/: Fix typos in doxygen comments. Sat Jun 28 10:33:06 GMT 2008 Olly Betts <olly@survex.com> * languages/stem.cc: Add "nl" for selecting the Dutch stemmer, which was accidentally missing from the list of language codes recognised. Reported by Joey Hess in Debian bug #484458. * tests/api_nodb.cc: Extend stemlangs1 to check that all the language codes work, and so does Stem(""). Fri Jun 27 00:27:31 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog and against 1.0 branch. Thu Jun 26 17:17:41 GMT 2008 Olly Betts <olly@survex.com> * net/remoteconnection.cc: The "already done" check in do_close() is no longer useful, so remove it. Shrink the try block to only cover the call to send_message(). Thu Jun 26 08:19:41 GMT 2008 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Don't bother to send MSG_SHUTDOWN for a read-only Database - just closing the connection is enough. Thu Jun 26 04:49:06 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_btreebase.cc: Don't need <string.h> as well as <cstring>. * backends/flint/flint_btreebase.cc,bin/xapian-replicate-server.cc: Prefer <cXXX> to <XXX.h>. Tue Jun 24 04:31:10 GMT 2008 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Use resize() to truncate a string. Tue Jun 24 04:23:59 GMT 2008 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Fill in a guess at the __WIN32__ version of the code needed in RemoteConnection::do_close(). Tue Jun 24 04:06:02 GMT 2008 Olly Betts <olly@survex.com> * net/progclient.cc: Fix debug logging for __WIN32__. Tue Jun 24 03:11:09 GMT 2008 Olly Betts <olly@survex.com> * net/remoteserver.cc: Just delete the Database * pointer db. The Database dtor is virtual, so it's fine to delete a WritableDatabase via a Database * pointer. Tue Jun 24 01:29:12 GMT 2008 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remoteconnection.h, net/remoteconnection.cc,net/replicatetcpclient.cc: Wait for the connection to close rather than using a different shutdown message which requires a reply. Mostly this is more backport-friendly, but it also avoids a message reply (albeit it not in a performance sensitive situation). Needs implementing for __WIN32__. Tue Jun 24 01:07:16 GMT 2008 Olly Betts <olly@survex.com> * net/remoteserver.cc: Unwrap comment. Tue Jun 24 00:52:02 GMT 2008 Olly Betts <olly@survex.com> * net/remoteserver.cc: Kill a blank line. Tue Jun 24 00:49:50 GMT 2008 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remoteconnection.h, common/remoteprotocol.h,docs/remote_protocol.html, net/remoteconnection.cc,net/remoteserver.cc, net/replicatetcpclient.cc: Revert r10713 in preparation for applying an alternative fix. Mon Jun 23 21:51:42 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Expand list of keywords for brace cuddling. Add rationale. Mon Jun 23 01:53:03 GMT 2008 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Cuddle braces after for loops for consistency with the rest of the code. Use string::resize() to shrink a string rather than assigning the result of substr(0, x). * HACKING: Explicitly document that braces should be cuddled after control flow structures. Mon Jun 23 01:34:39 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/perftest/perftest_matchdecider.cc: Reworked performance test to include a test of an alldocspostingiterator, and to only rebuild the database involved if it's not up-to-date. Mon Jun 23 01:20:25 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/valuestats.h: Add missing #include. Mon Jun 23 01:10:19 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-compact.cc: Add support for the valuestats which are now held in the value table for chert. Sun Jun 22 23:44:46 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/: Change ChertAllDocsPostList to inherit from ChertPostList and use the doclen list from the posting table rather than using the termlist table. This helps towards making the termlist table optional, reduces the amount of data read in the process of iterating through an alldocs postlist, and can make a massive difference in performance: I've measured the time to iterate through all the documents in a 1000000 document database, and the patch speeds this operation up by a factor of 6. Add `keep_reference` parameter to the ChertPostList constructor, instead of not keeping a reference if the term is empty. ChertAllDocsPostList uses this to keep a reference to the database while using an empty term. Add ChertAllDocsModifiedPostList class, inspired by ChertModifiedPostList, (with corresponding new source files) to handle alldocs postlist with modifications: this wasn't needed before since the termlist is updated immediately after changes. Sun Jun 22 21:37:00 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/harness/: Add "name" parameter to BackendManager::get_writable_database_as_database(), and BackendManager::get_writable_database_again(), (and subclasses) so that old databases can be revived. Useful for the performance tests. Sun Jun 22 09:30:44 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/api_anydb.cc: Expand tests for alldocspostlist, covering iteration of modified databases. Fri Jun 20 08:28:12 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/remote/remote-database.cc,common/remoteconnection.h, common/remoteprotocol.h,docs/remote_protocol.html, net/remoteconnection.cc,net/remoteserver.cc, net/replicatetcpclient.cc: Add new message "MSG_SHUTDOWNANDCONFIRM" which causes a reply of "REPLY_SHUTDOWNCONFIRMATION" to be sent after the database has been closed. Use this message when closing a writable database, to ensure that the destructor doesn't return until the lock on the database has been released. Tue Jun 17 14:28:46 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/perftest/freemem.cc,tests/perftest/perftest.cc: Patches from Charlie to allow perftest to compile on windows. Tue Jun 17 13:23:18 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_btreebase.cc,backends/chert/chert_lock.cc, backends/flint/flint_btreebase.cc,backends/flint/flint_lock.cc, bin/xapian-replicate-server.cc: Add some missing includes, needed for GCC 4.3. Tue Jun 17 11:00:56 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/perftest/get_machine_info.in: Updates to hopefully get useful version information on windows. Mon Jun 16 12:13:48 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/harness/testsuite.cc: Don't call backendmanager->posttest() if no backendmanager is in use. Sun Jun 15 02:19:03 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/harness/: Wait for subprocesses to finish at end of tests, with remotetcp backend, to avoid test failures due to final flush of a subprocess writing to the database used for the next test. Thu Jun 12 09:09:44 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/unaligned.h: Add some casts to make GCC 4.3.0 happy. Thu Jun 12 09:06:21 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_btreebase.cc: Add missing #include (needed for compilation with GCC 4.3.0). Tue Jun 10 17:38:07 GMT 2008 Olly Betts <olly@survex.com> * tests/harness/testrunner.cc: Use startswith(). Fix memory leak. Fri Jun 06 11:02:56 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * Makefile.am,tests/Makefile.am: Fix make check-* targets to match the current set of backends. Add .multichert and .multiflint to the clean-local hook in tests/. Leave .multi there for now, to clean up existing .multi directories in the automated build trees. Fri Jun 06 08:26:49 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/Makefile.mk: Add valuestats.h to distribution. Thu Jun 05 23:37:37 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_values.cc: Fix compile error in assertion in recently added code for calculating value statistics. Thu Jun 05 21:38:01 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Use std::vector<Term *> rather than std::list<Term *> so that size() is O(1) with any (sane) compiler and because a list of pointers isn't space efficient (and we don't need to splice here). Factor out common code in if branches in TermGroup::as_group(). Thu Jun 05 17:04:23 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Use "startswith()" to avoid creating a new string object. Thu Jun 05 09:09:54 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Remove variable which is set but never otherwise used. Thu Jun 05 09:08:22 GMT 2008 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Don't call find() twice - reuse the result we just got. Use AssertEq() rather than Assert() on an equality test. Clarify a comment. Wed Jun 04 21:32:11 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.lemony: Fix more O(N*N) scaling, this time in the way in which auto multiword synonyms are detected. Instead of trying all possible sub-sequences of terms for synonyms, use synonym_keys and the skip_to() method to check for synonyms which actually exist in the synonyms table. * tests/queryparsertest.cc: Extend the scaling test to check this case. Wed Jun 04 12:42:32 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.lemony: Fix various cases where queries were constructed pair-wise within a loop, which leads to O(N*N) scaling behaviour (because each intermediate query construction is O(M) where M is the size of that query, and there are N of them). * tests/queryparsertest.cc: Add test to check that parsing of a query scales roughly linearly with query size. * common/omtime.h: Add method to get an OmTime as a double - used by the new query parser test. Wed Jun 04 12:18:35 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/harness/testsuite.h: Fix TEST_LESSER_OR_EQUAL and TEST_LESSER macros to compare in the right direction (these macros are, so far, unused). Mon Jun 02 18:19:22 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc: Make ValueWeightPostingSource use value statistics, if they're available, to return accurate term frequency statistics, and a better maxweight. * tests/api_db.cc: Add valueweightsource2, to test the bounds and maxweight returned by a ValueWeightPostingSource for backends with valuestats support. Replace !multi condition in valueweightsource1 with a SKIP_TEST_FOR_BACKEND invocation, since this is a shortcoming we should fix at some point. Mon Jun 02 18:16:52 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/harness/testrunner.cc,tests/harness/testrunner.h: Replace DO_TESTS_FOR_BACKEND macro with a do_tests_for_backend method; the method version takes a BackendManager * and gets the backend name from that, rather than requiring it to be specified. Requires addition of a couple of extra private members to the TestRunner class. Mon Jun 02 10:27:45 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * include/xapian/database.h: Add get_value_freq(), get_value_lower_bound() and get_value_upper_bound() methods to Database, to get statistics about the values stored in a slot. * api/omdatabase.cc,backends/chert/,backends/database.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/remote/remote-database.cc,common/, net/remoteserver.cc: Add support for the value statistics methods to chert, inmemory, multi and remote databases. * tests/Makefile.am,tests/api_valuestats.cc, tests/harness/testrunner.cc,tests/harness/testrunner.h: Add test of the value statistics code, and a "valuestats" backend property to select only backends supporting this. Mon Jun 02 10:26:12 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/harness/backendmanager.cc,tests/harness/backendmanager.h: Move implementation of get_dbtype() method out of header. Mon Jun 02 09:44:34 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * rests/harness/: Make flint and chert subtypes for the remote and multi backend types. These are specified (on the command line and elsewhere) by the main type (multi, remotetcp or remoteprog), followed by an underscore, followed by the subtype (flint or chert). Change return type of BackendManager::get_dbtype() to a string instead of a const char *, to allow backend managers which have subtypes to generate the result dynamically. * tests/harness/backendmanager_remote.cc, tests/harness/backendmanager_remote.h, tests/harness/backendmanager_remoteprog.cc, tests/harness/backendmanager_remoteprog.h, tests/harness/backendmanager_remotetcp.cc, tests/harness/backendmanager_remotetcp.h: Add new BackendManagerRemote class as a common base for the two remote backend managers, and move duplicated code into it. Add parameter to its constructor, and to the constructors of the remote backend manager subclasses, to control the type of database to use at the remote end. * tests/harness/backendmanager_multi.cc, tests/harness/backendmanager_multi.h: Add parameter to constructor of BackendManagerMulti, controlling the type of database to use for sub databases. Put the generated database files in ".multiflint" and ".multichert" instead of ".multi". * tests/harness/testrunner.cc: Update list of backend properties to contain the new subtypes. Modify use_backend() to allow either the full backend type, or just the main part of the backend type, to be specified, so that, for example "-b multi" will run both the multi_chert and multi_flint tests. Add test runs for all the subtypes now supported. Catch string exceptions thrown by the test runner - these can currently occur when neither flint or chert are compiled in. * tests/apitest.cc,tests/apitest.h: Adjust return type of get_dbtype(), implement skip_test_*_backend functions which just check the prefix of the type, and implement SKIP_TEST_*_BACKEND macros using these. * tests/api_anydb.cc: Modify check for running on a multi backend to copy with getting a string back from get_dbtype(). Sat May 31 00:55:03 GMT 2008 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/perftest/Makefile.mk: Remove unnecessary quotes. Thu May 29 16:04:56 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/perftest/: Update svn:ignore property for perftest changes. Thu May 29 14:33:51 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/perftest/perftest.cc: Include the xapian version string in the test output. For runs from SVN, also include the subversion revision number and branch in the output. Wed May 28 21:47:37 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/valuegepostlist.cc,matcher/valuerangepostlist.cc, matcher/valuerangepostlist.h: Apply patch from ticket #270 to implement ValueRangePostList::next() by using an alldocs posting list, instead of trying docids in turn. This is much more efficient if document IDs are sparse. Since both methods require accessing the termlist table to check for the next document, and some cursory performance tests don't show a measurable difference in speed, I think this approach is at least safe, and probably superior. Wed May 28 21:28:38 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/perftest/perftest.cc: Log the flush threshold for indexing runs. * tests/perftest/perftest_randomidx.cc: Set the run size back to a reasonably large value. Wed May 28 20:48:10 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * configure.ac,tests/,tests/perftest/,tests/perftest_matchdecider.cc, tests/perftest_randomidx.cc: Move performance test source files into a subdirectory. Modify output of performance tests to include some basic system information, and also to include details of the parameters used to perform indexing tests. Wed May 28 20:42:56 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * HACKING: Fix numbering in checklist for developers so that it doesn't repeat section 5. Sun May 25 14:21:40 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update to match 1.0 branch. Sun May 25 00:24:09 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Start to update from ChangeLog, stripping out backported changes. Sat May 24 16:38:49 GMT 2008 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update to match 1.0 branch. Sat May 24 16:33:19 GMT 2008 Olly Betts <olly@survex.com> * tests/termgentest.cc: Pass Xapian::Document by const reference. Sat May 24 14:30:59 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Only warn about duplicate user metadata keys if the tags aren't the same. Sat May 24 12:14:55 GMT 2008 Olly Betts <olly@survex.com> * bin/Makefile.am: Put special -I options in foo_CPPFLAGS not foo_CXXFLAGS. As well as being the more correct place, this also means that AM_CXXFLAGS is now used when compiling xapian-check and xapian-compact, which in particular enables compiler warnings. * bin/xapian-check.cc,bin/xapian-compact.cc,bin/xapian-check-flint.cc: Fix warnings (one unused variable and a few cases of a variable masking another with the same name - none problematic in practice). Fri May 23 09:22:24 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/inmemory/inmemory_database.cc: Fix bug in inmemory database, which resulted in the values not being stored correctly if document IDs were sparse. * matcher/valuerangepostlist.cc: Fix the next() method not to return document IDs which aren't present in the database - this used to happen if the document IDs were sparse. * tests/api_anydb.cc: Add "valuerange2", a regression test for OP_VALUE_LE returning document IDs which aren't present in the database, and "alldocspl1", which does a generic test of an alldocs postlist in a sparse database (this passed before the above changes, but might as well be added anyway). Wed May 21 19:33:12 GMT 2008 Olly Betts <olly@survex.com> * docs/Makefile.am: test == isn't portable - use test = instead. Wed May 21 13:22:23 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Fix path for atreus. Wed May 21 13:21:08 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Fix search&replace error - the CVS module is still called www.xapian.org. Wed May 21 13:16:39 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Reword, since environmental variable PATH isn't "set by $PATH". Wed May 21 13:16:21 GMT 2008 Olly Betts <olly@survex.com> * HACKING: ixion -> atreus. Wed May 21 13:12:07 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/database.h: Improve documentation comment. Sun May 18 05:20:29 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * docs/remote_protocol.html: Remove spurious "i" at end of line. Sat May 17 11:53:54 GMT 2008 Olly Betts <olly@survex.com> * tests/harness/testutils.h: Layout and macro parameter name tweaks. Fri May 16 20:42:05 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Check that all source databases are the same type (flint or chert). Copy over "iamchert" for chert. Update terminology - the "meta file" is now called the "version file". Fri May 16 17:33:22 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/chert/chert_version.cc: Need <cstdio> for rename(). Fri May 16 17:27:56 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc: Correct FIXME comment and mark as "FIXME:1.2.0". Fri May 16 17:21:59 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/harness/testrunner.cc,tests/harness/testrunner.h: Add condition to test if the backend is "inmemory". * tests/perftest_randomidx.cc: Change the run size from 1000 documents to 1000000. Disable this test for the inmemory backend, because it will quickly use up all the memory on the machine. Fri May 16 16:56:25 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * HACKING: Add note about some more PDF generation tools which are needed, and that doxygen requires them to be on PATH. Fri May 16 16:38:40 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Ubuntu now needs tetex-extra too. Fri May 16 10:05:30 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Fix to not reject chert databases during the "up front" check. Thu May 15 20:20:14 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * net/progclient.cc: Set namespace for close methods (to avoid conflicting with the close method I'm working on adding to database internal). Wed May 14 17:14:53 GMT 2008 Olly Betts <olly@survex.com> * backends/database.cc: Default to flint not chert for a new database. Wed May 14 17:11:56 GMT 2008 Olly Betts <olly@survex.com> * api/postingsource.cc,docs/postingsource.rst, include/xapian/postingsource.h,matcher/externalpostlist.cc: PostingSource::check() now returns the valid flag rather than passing it in by reference to be set. Change "should" to "must" for the get_termfreq_est() requirement at Richard's suggestion. Add parameter names to PostingSource method prototypes and update documentation comments to match. Wed May 14 15:34:38 GMT 2008 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/index.html,docs/postingsource.rst: Add topic document for PostingSource. Mon May 12 17:27:13 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * include/xapian/postingsource.h: Add documentation comments for ValueWeightPostingSource, and tweak PostingSource::get_maxweight() comment a bit more. Mon May 12 16:51:01 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/postingsource.h: Clarify get_maxweight() if the maximum from now on. Mon May 12 06:51:53 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/postingsource.cc,include/xapian/postingsource.h,tests/api_db.cc, tests/harness/index_utils.cc: Add ValueWeightPostingSource, which reads a value, applies sortable_unserialise() to it, and returns that as the weight for each document. Sun May 11 23:49:07 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Do a final sync by comparing our versions to the latest vanilla upstream. Add a note to each file of the latest revision sync-ed against. * queryparser/lemon.c: Enable '#define PRIVATE static'. Sun May 11 22:48:27 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Add a note of the speed-up that the recent QueryParser changes have given. Sun May 11 22:10:19 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 5053". Sun May 11 22:01:48 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 5052". Sun May 11 21:56:03 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Fix a comment typo. Sun May 11 21:52:30 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 4751". Sun May 11 21:51:18 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 4745". Sun May 11 21:47:03 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Comment out memset() call added by one of the earlier merged upstream changes - a later comment on the ticket referred to notes that this was actually a bug in the grammar of the people who reported it. Sun May 11 21:44:46 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 4743". Sun May 11 21:41:55 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 4738". Sun May 11 21:33:13 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Merge upstream "Check-in Number: 4736". * queryparser/lemon.c: Add %syntax_error directive since lemon has changed it's error recovery strategy if you don't have one. Sun May 11 20:54:31 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: If we hit a syntax error, don't bother feeding further tokens to the parser! Sat May 10 21:42:44 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 4641". Sat May 10 21:33:10 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 4473". Sat May 10 21:31:25 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 4439". Sat May 10 21:28:27 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 4274". Sat May 10 21:25:35 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 4190". Sat May 10 21:20:16 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 4160". Sat May 10 21:07:15 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 3753". Sat May 10 21:03:19 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Merge upstream "Check-in Number: 3738". Sat May 10 20:00:09 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 3654". Sat May 10 19:55:32 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 3594". Sat May 10 19:46:05 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 3593". Sat May 10 19:44:15 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 3591". Sat May 10 19:40:12 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 3528". Sat May 10 19:38:36 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 3476". Sat May 10 19:36:06 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 3333". Sat May 10 19:33:33 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 3244". Sat May 10 19:29:18 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 3226". Sat May 10 19:19:48 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 3224". Sat May 10 19:14:18 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Merge upstream "Check-in Number: 3210". Sat May 10 18:56:50 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 3181". Sat May 10 17:05:35 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 3126". Sat May 10 16:51:31 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 2764". Sat May 10 16:40:48 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c,queryparser/queryparser.lt: Merge upstream "Check-in Number: 2761". Sat May 10 16:32:45 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 2458". Sat May 10 16:29:55 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 2345". Sat May 10 16:08:28 GMT 2008 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Merge upstream "Check-in Number: 2302". Sat May 10 15:50:23 GMT 2008 Olly Betts <olly@survex.com> * queryparser/lemon.c: Merge upstream "Check-in Number: 2208". Sat May 10 15:03:59 GMT 2008 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Prefer `str.assign(ptr)' to `str = string(ptr)'. Sat May 10 14:48:38 GMT 2008 Olly Betts <olly@survex.com> * api/matchspy.cc,api/replication.cc: Prefer `str.assign(ptr, len)' to `str = string(ptr, len)'. Fri May 09 16:53:19 GMT 2008 Olly Betts <olly@survex.com> * net/remoteserver.cc: No entries in dispatch[] are NULL, so there's no point testing for it. Fri May 09 16:51:10 GMT 2008 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Remove pointless conversion to size_t which breaks GCC 2.95 build. Fri May 09 16:22:46 GMT 2008 Olly Betts <olly@survex.com> * api/replication.cc: Needs <cstdio> for rename(). Fri May 09 15:41:08 GMT 2008 Olly Betts <olly@survex.com> * tests/harness/testutils.cc: No need for an explicit std::endl when calling TEST_AND_EXPLAIN(). Fri May 09 15:38:22 GMT 2008 Olly Betts <olly@survex.com> * api/replication.cc: Use '\n' rather than endl in the middle of writing stuff, since endl forces a flush which just adds useless overhead. Fri May 09 15:24:42 GMT 2008 Olly Betts <olly@survex.com> * common/omdebug.h: Fix RETURN macro to work on GCC 2.95 for the case `RETURN(string())'. Fri May 09 08:33:35 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/postingsource.h: Add a protected default ctor so that code subclassing PostingSource actually compiles! Thu May 08 16:41:59 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/postingsource.h: Add doxygen comments for Xapian::PostingSource. Add private assignment operator and copy constructor to prevent copying. Thu May 08 16:40:49 GMT 2008 Olly Betts <olly@survex.com> * common/postlist.h: Fix one doxygen comment and improve another. Mon May 05 15:22:24 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am: Clean up .chert in clean-local. Mon May 05 14:52:34 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * docs/Makefile.am: Adapt the dist-check hook to work with a VPATH build in non-maintainer mode. Mon May 05 13:18:19 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * docs/Makefile.am: Fix "make distcheck" by using dist-hook to install generated files, with the appropriate dependency, instead of wildcards in EXTRA_DIST which don't give the correct expansion unless the files happened to be generated already by another rule. Also, ensure that the documentation is generated before attempting to install it. Sun May 04 09:30:53 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_version.cc,backends/flint/flint_version.cc, common/stringutils.h: Move CONST_STRLEN() into stringutils.h. * backends/chert/chert_database.cc,backends/flint/flint_database.cc: Use CONST_STRLEN(). Sat May 03 18:56:34 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omenquire.cc: Reset the internal sorter pointer to NULL when changing to a sort by value. Fixes #256. * tests/api_sorting.cc: Add regression test. Sat May 03 18:52:27 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Mark internal functions as @internal. Sat May 03 18:43:12 GMT 2008 Olly Betts <olly@survex.com> * docs/doxygen_api_conf.in: Disable header and directory relationship graphs in the API documentation as they aren't interesting to users. Sat May 03 18:12:30 GMT 2008 Olly Betts <olly@survex.com> * docs/gen_codestructure_doc.in: Link to trac instead of viewvc. Sat May 03 17:25:02 GMT 2008 Olly Betts <olly@survex.com> * docs/replication_protocol.rst: Fix a few typos. Sat May 03 15:53:24 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_database.cc,backends/flint/flint_database.cc: If Database::reopen() is called and the database revision on disk hasn't changed, then do as little work as possible. Even if it has changed, don't bother to recheck the version file (bug#261). Sat May 03 10:28:30 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: Improve code to prevent probing for f77, etc. Sat May 03 10:17:05 GMT 2008 Olly Betts <olly@survex.com> * matcher/rset.cc: Missing change from last commit. Sat May 03 09:25:29 GMT 2008 Olly Betts <olly@survex.com> * api/omenquire.cc,common/,expand/Makefile.mk,expand/esetinternal.cc, expand/expand.cc: Rewrite class ESet::Internal, incorporating the functionality which used to be in the internal OmExpand class. The expand operation now uses a min heap rather than calling nth_element() repeatedly - this should reduce the complexity of the expand operation by a factor of n/log(n) where n is the requested ESet size. Fri May 02 15:51:29 GMT 2008 Olly Betts <olly@survex.com> * common/ortermlist.h,expand/ortermlist.cc: Kill FreqAdderOrTermList's explicit empty dtor. Thu May 01 16:38:40 GMT 2008 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,api/replication.cc,backends/chert/, backends/flint/,backends/inmemory/inmemory_positionlist.cc, common/utils.cc,matcher/stats.cc,tests/perftest_matchdecider.cc, tests/perftest_randomidx.cc: Assorted formatting tweaks. Thu May 01 16:32:58 GMT 2008 Olly Betts <olly@survex.com> * common/unaligned.h: Fix ";;" to ";" after AssertRel calls. Thu May 01 16:12:05 GMT 2008 Olly Betts <olly@survex.com> * docs/doxygen_api_conf.in,docs/doxygen_full_conf.in: Tell doxygen to expand XAPIAN_VISIBILITY_DEFAULT so it doesn't appear in collated API documentation. Wed Apr 30 10:57:53 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Update bugzilla references and URLs to reflect the move to trac. Update the "fixing a bug" checklist to mention backporting and updating the release notes. Wed Apr 30 10:40:30 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/remote/remote-database.cc: Fix error message displayed when remote protocol version doesn't match to display the minor part of the version number supplied by the server correctly. Tue Apr 29 17:37:10 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * AUTHORS: Add Alexandre Gauthier, for supplying a method for doing python packaging for Windows. Tue Apr 29 16:20:47 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/fileutils.cc: Fix variable name for windows compilation. Tue Apr 29 06:50:53 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_table.cc,backends/flint/flint_table.cc: Remove some debugging code, left over from writing the replication code. Mon Apr 28 12:11:38 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/chert/chert_btreebase.cc,backends/chert/chert_table.cc, backends/flint/flint_btreebase.cc,backends/flint/flint_table.cc: Fix issue #259 - update the last_block value whenever allocating a new value, so that next_for_sequential doesn't give up too early. Also, in both next_for_sequential and prev_for_sequential, for writable databases, check the built-in cursor at all levels for the new block rather than just reading from disk, since newly allocated blocks won't have been written to disk yet. Also, test the return type of next() when reading subsequent parts of tags, and raise DatabaseCorruptError if it returns false; this prevents incorrect data being returned, leading to a zlib error. * tests/api_wrdb.cc: Add regression test. Fri Apr 25 08:39:28 GMT 2008 Olly Betts <olly@survex.com> * api/postingsource.cc: Missed change from last but one commit. Thu Apr 24 13:44:48 GMT 2008 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Reserve size of postlists vector. Thu Apr 24 13:38:51 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/postingsource.h,matcher/externalpostlist.cc, matcher/externalpostlist.h,tests/api_db.cc: Add a default PostingSource::get_description() method so users who don't care aren't forced to define one. Make PostingSource::reset() a standard method and call it automatically so that a PostingSource-using Query can be run more than once without extra work. Thu Apr 24 11:24:30 GMT 2008 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Don't need <queue>. Thu Apr 24 04:50:43 GMT 2008 Olly Betts <olly@survex.com> * tests/api_db.cc: Add test that boolean branches of matches don't cause Xapian::PostingSource to be asked for weights. Thu Apr 24 03:27:39 GMT 2008 Olly Betts <olly@survex.com> * matcher/externalpostlist.cc,matcher/externalpostlist.h, matcher/queryoptimiser.cc: ExternalPostList now takes the factor into account, so works with OP_SCALE_WEIGHT, and doesn't call get_weight()/get_maxweight() for a boolean branch. * tests/api_db.cc: Add tests for OP_SCALE_WEIGHT. Thu Apr 24 02:16:08 GMT 2008 Olly Betts <olly@survex.com> * xapian-config.in: Back out --swigheaders. Wed Apr 23 11:37:45 GMT 2008 Olly Betts <olly@survex.com> * matcher/externalpostlist.h: Need to initialise current. Wed Apr 23 07:57:35 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/termgentest.cc: Put the inclusion of "utils.h" back; needed for om_tostring(). Wed Apr 23 04:20:55 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/replication.h: Eliminate reference to FlintDatabase from doc comment since it's not user-visible, and chert supports replication too. No need for '#include <xapian/database.h>' so remove. Remove references to non-existent parameter 'remotename'. 'DatabaseCorrupt error' -> 'DatabaseCorruptError'. Wed Apr 23 04:19:53 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/query.h: Wrap Xapian::Query::Internal class definition in "#ifndef SWIG" ... "#endif" in preparation for getting SWIG to directly parse it. Wed Apr 23 02:16:53 GMT 2008 Olly Betts <olly@survex.com> * api/omquery.cc: Prefer string() to "". Wed Apr 23 02:16:10 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/query.h: Fix doc comment typo (doesn't affect API docs). Wed Apr 23 00:59:40 GMT 2008 Olly Betts <olly@survex.com> * api/postingsource.cc: Another missing file. Wed Apr 23 00:34:34 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/postingsource.h: New file missing from previous commit. Wed Apr 23 00:30:34 GMT 2008 Olly Betts <olly@survex.com> * api/Makefile.mk,api/omquery.cc,api/omqueryinternal.cc, include/Makefile.mk,include/xapian.h,include/xapian/query.h,matcher/, tests/api_db.cc: Add support for Xapian::PostingSource. Wed Apr 23 00:21:25 GMT 2008 Olly Betts <olly@survex.com> * include/xapian.h: Fix misplaced comment. Tweak doxygen comments for version functions to be more consistent with others. Wed Apr 23 00:16:14 GMT 2008 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Remove bogus comment from cut-and-paste. Wed Apr 23 00:14:30 GMT 2008 Olly Betts <olly@survex.com> * common/omassert.h: Add AssertEqDoubleParanoid(). * matcher/multimatch.cc: Use it instead of home-brewed version. Tue Apr 22 23:30:50 GMT 2008 Olly Betts <olly@survex.com> * common/utils.h: Stop exporting internal functions needlessly. Tue Apr 22 12:53:55 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Fix to work again (broken by the splitting of FlintTable::commit() into flush_db() and commit() in the replication changes). Don't bother passing tablename to functions where it always has the same value. Tweak the user-metadata copying loop to make it clearer. Reserve the right size for the vector used for merging spellings. Tue Apr 22 12:28:17 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_table.h,backends/flint/flint_table.h, bin/xapian-compact.cc: Default commit's changes_fd parameter to -1. Tue Apr 22 11:39:10 GMT 2008 Olly Betts <olly@survex.com> * common/omassert.h: Fix comment typo. Tue Apr 22 10:54:35 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/queryparsertest.cc: Add "utils.h" back - needed for om_tostring() and mkdir(). Mon Apr 21 07:48:45 GMT 2008 Olly Betts <olly@survex.com> * tests/api_db.cc: Don't need <list>. Wrap comment. Mon Apr 21 07:04:12 GMT 2008 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,backends/chert/chert_check.cc, backends/flint/flint_check.cc,tests/: Remove various needless inclusions of headers (especially <iostream>). Reorder headers. Wrap a couple of comments. Append single characters to strings using character constants. Mon Apr 21 06:26:49 GMT 2008 Olly Betts <olly@survex.com> * api/Makefile.mk,api/valuerangeproccompat.cc, include/xapian/queryparser.h: Remove the v102::NumberValueRangeProcessor ABI-preserving machinery since it's OK to change the ABI for 1.1.0. Mon Apr 21 04:42:52 GMT 2008 Olly Betts <olly@survex.com> * api/,queryparser/queryparser.cc: Remove debug log tracing from get_description() methods since the debug log call tracing *calls* get_description() methods on parameters, so logging these calls just makes for more confusing debug logs. A get_description() method should have no side-effects so it's not very interesting even when explicitly called by the user. * common/omdebug.h: Replace "OM_DEBUG_INTRO" class with dummy placeholder (to preserve numbering), to help prevent accidentally adding these back. Sun Apr 20 09:46:11 GMT 2008 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,common/: Stop describing get_description() as an "Introspection method" internal (matching earlier change to stop doing so externally). This doesn't help to explain what it does, and get_description() doesn't actually fall under any of the formal definitions of "introspection" I can find. Fri Apr 18 10:26:18 GMT 2008 Olly Betts <olly@survex.com> * backends/chert/chert_metadata.cc: Fix renamed parameter in debug logging. Fri Apr 18 10:18:35 GMT 2008 Olly Betts <olly@survex.com> * HACKING,INSTALL,Makefile.am,backends/Makefile.mk,backends/chert/, backends/database.cc,backends/flint/,backends/remote/net_postlist.cc, backends/remote/net_postlist.h,bin/,common/,configure.ac,docs/, include/xapian/dbfactory.h,include/xapian/version_h.cc, net/remoteserver.cc,tests/Makefile.am,tests/harness/: Start new development backend "chert". Currently the change over flint is that in the postlist table, doclengths are stored once in a chunked postlist-like form, rather than once per posting. Tue Apr 15 23:24:39 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omenquire.cc,tests/api_nodb.cc: Fix segfault introduced by previous change when asking for the termfreq on an empty mset. Include a regression test. Tue Apr 15 06:02:12 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * include/xapian/enquire.h: Change definition of MSet::get_termfreq such that it will fall back to looking the term frequency up in the database rather than raising an exception if the term is not present in the mset. * api/omenquire.cc,common/omenquireinternal.h: Implementation of the above. * tests/api_anydb.cc: Test the above. Mon Apr 14 21:29:59 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/: Use copies of freemem.cc and freemem.h from omega to report the free physical memory at each stage of indexing. Also, add a get_total_physical_memory() function to this, and report it's value at the top of the results file (only tested on Linux so far). Mon Apr 14 19:17:22 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/api_wrdb.cc: Make skip_to() test a bit more comprehensive. Mon Apr 14 19:08:49 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_metadata.cc,backends/flint/flint_metadata.h: Fix a bug in the skip_to() implementation, and set the parameter name to "key" instead of "tname", since it's not a term. * tests/api_wrdb.cc: Add a test of the metadata iterators. Mon Apr 14 14:22:16 GMT 2008 Olly Betts <olly@survex.com> * backends/Makefile.mk,backends/flint/Makefile.mk, backends/flint/contiguousalldocspostlist.cc, backends/flint/contiguousalldocspostlist.h,common/Makefile.mk: The ContiguousAllDocsPostList class isn't flint-specific, so move it out of backends/flint. Mon Apr 14 13:56:51 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Removed duplicate '#include "flint_database.h"', ironically added by Richard's recent change. Fix a couple of errors in alphabetical ordering, and move autoptr.h and the "safe" headers to where the standard headers they replace would go. Mon Apr 14 13:44:37 GMT 2008 Olly Betts <olly@survex.com> * Makefile.am,tests/Makefile.am: List check-perf in .PHONY. * HACKING: Document the need to list non-file targets in .PHONY. Mon Apr 14 07:59:56 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc: Correct return type in a DEBUGAPICALL macro. Mon Apr 14 07:14:01 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc,backends/database.cc,backends/flint/, common/database.h,include/xapian/database.h: Add metadata_keys_begin() and metadata_keys_end() methods to Database, to allow the complete list of metadata in a database to be retrieved. Supports a prefix argument to obtain a limited subset of the keys, since that may be useful in some situations and is very easy to implement. Only accesses the keys in the first database in a multidatabase situation (which corresponds with the current behaviour of get_metadata()). Also, reorder the includes in flint_database.cc to be largely alphabetical, and remove a duplicated #include of <xapian/error.h>. Mon Apr 14 06:51:48 GMT 2008 Olly Betts <olly@survex.com> * common/fileutils.cc: Use endswith(). Mon Apr 14 06:14:44 GMT 2008 Olly Betts <olly@survex.com> * api/replication.cc,backends/database.cc, backends/flint/flint_database.cc,common/fileutils.cc, tests/queryparsertest.cc: Use empty() rather than comparing size() to 0. Use startswith() in a couple of places. Tweak an Assert to AssertEq. Sun Apr 13 14:30:55 GMT 2008 Olly Betts <olly@survex.com> * api/omdatabase.cc: Replace uses of DEBUGLINE(SPELLING, ...) in Database::get_spelling_suggestion() with standard DEBUGAPICALL/RETURN tracing. Add DEBUGAPICALL/RETURN tracing to a few Database methods which were missing it. Sun Apr 13 14:15:00 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.cc: Use RETURN() in a few places which weren't. Sun Apr 13 13:59:57 GMT 2008 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc: Remove long-time commented out uses of DebugMsg. Sun Apr 13 13:52:32 GMT 2008 Olly Betts <olly@survex.com> * expand/expandweight.cc: Use DEBUGLINE() instead of DEBUGMSG(). Sun Apr 13 11:23:56 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Include corresponding header right after config.h. Sun Apr 13 10:28:27 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Tweak brace formatting and wrap a comment. Fri Apr 11 16:18:14 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/fileutils.cc: Fix typo in windows specific code. Fri Apr 11 12:44:25 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.cc: Include corresponding header right after config.h. Fri Apr 11 10:25:03 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.h: Fix filename in header comment. Fri Apr 11 04:49:03 GMT 2008 Olly Betts <olly@survex.com> * configure.ac,tests/submitperftest.in,tests/submitperftest.py.in: Rename submitperftest.py to submitperftest. * HACKING: Document that scripts shouldn't generally have an extension, and the reason for this policy. Fri Apr 11 04:32:15 2008 Olly Betts <olly@survex.com> * HACKING: Note that GCC 4.3 no longer supports pre-ISO forms of standard headers, such as <list.h>. GCC on Solaris now seems to support throwing exceptions across shared library boundaries. Fri Apr 11 03:24:47 GMT 2008 Olly Betts <olly@survex.com> * tests/Makefile.am: "make up" in tests now does "make" in the top-level. Thu Apr 10 08:33:06 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_postlist.cc: Fix debug logging message: ostList -> FlintPostList. Thu Apr 10 05:47:37 GMT 2008 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Remove erroneous FIXME comments talking about ReadFile() not updating WSAOVERLAPPED's Offset/OffsetHigh being a problem with using _get_osfhandle() - ReadFile is documented to behave this way on MSDN. Factor out update of Offset/OffsetHigh and tweak this code not to rely on off_t being 64 bits. Thu Apr 10 05:24:11 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_record.cc,backends/flint/flint_utils.h: Use the new STATIC_ASSERT family of macros instead of CASSERT and CASSERT_TYPE_UNSIGNED. Thu Apr 10 04:46:40 GMT 2008 Olly Betts <olly@survex.com> * common/omassert.h: Add "compile-time assertion" macros: STATIC_ASSERT(COND), STATIC_ASSERT_UNSIGNED_TYPE(TYPE), STATIC_ASSERT_TYPE_DOMINATES(TYPE1, TYPE2). * tests/internaltest.cc: Test these. Thu Apr 10 01:07:34 GMT 2008 Olly Betts <olly@survex.com> * common/remoteprotocol.h,net/remoteserver.cc: Clean up a couple more things now we've moved to major protocol version 31. Wed Apr 09 06:50:44 GMT 2008 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc: Add missing explicit include of remote-database.h. Wed Apr 09 04:48:54 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-check.cc,bin/xapian-compact.cc: Fix check for user metadata key to not match other key types we may add in the future. When compacting, we can't assume how we should handle them. When checking, they currently shouldn't be present, so we should flag an error for them. Tue Apr 08 15:27:09 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Fix missing line in old entry. Mon Apr 07 07:49:56 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_btreeutil.h: Use AssertRel rather than Assert. Mon Apr 07 06:20:42 GMT 2008 Olly Betts <olly@survex.com> * api/error.cc,api/valuerangeproc.cc,backends/flint/flint_table.cc, backends/multi/multi_postlist.cc,backends/remote/net_termlist.cc, bin/xapian-inspect.cc,common/omdebug.cc,common/utils.cc, languages/stem.cc,languages/steminternal.cc, tests/harness/backendmanager_multi.cc, tests/harness/backendmanager_remotetcp.cc: Header inclusions tidying - remove a few redundant inclusions; prefer cstdio to stdio.h, etc in a few places; note why certain headers are required in cases where it's less obvious; in a .cc file, include the corresponding .h second (after <config.h>). Mon Apr 07 04:23:30 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: Fix --enable-log=profile to be recognised. * common/omdebug.cc,common/omdebug.h: Fix build with --enable-log=profile. * HACKING: Actually document --enable-log=profile. Sat Apr 05 21:07:38 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/perftest.cc: Flush after each write, so we can monitor progress of long running tests (we might need to remove this again later, if profiling shows it has a noticeable impact). Fix two calls to write() to be a single call. * tests/perftest_randomidx.cc: Add values to different slots; I'd added them all to slot 0 by mistake, so they were overwriting each other. Sat Apr 05 20:17:43 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am,tests/perftest_randomidx.cc: Add a new performance test which builds a randomly generated index, and times the index run. Probably wants a bit more tweaking to make it slightly more representative, but should be a useful test to run on systems where downloading a large amount of data is impractical. * ChangeLog: Add missing log message for last commit. Sat Apr 05 20:16:58 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/perftest.cc,tests/perftest.h: Log indexing runs every 5 seconds, as well as after every 1000 documents - gives useful information for slow index runs. Thu Apr 03 20:58:31 GMT 2008 Olly Betts <olly@survex.com> * backends/database.cc,include/xapian/dbfactory.h, include/xapian/version_h.cc: Remove lingering traces of quartz. Thu Apr 03 07:59:53 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-inspect.cc: Print top-bit-set characters as escaped hex forms as they often won't be valid UTF-8 sequences. Thu Apr 03 03:53:37 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-inspect.cc: Check for the user passing a database directory and issue a special error message since this is an obvious mistake to make. Wed Apr 02 07:00:36 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: Modify the estimate for the number of hits based on the rate at which a match decider has been denying documents. Also, reduce the upper bound based on the number of documents denied. * tests/api_db.cc: Check that the upper bound when a match decider is used is within the valid range, rather than a particular value. Tue Apr 01 22:40:19 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * configure.ac,tests/perftest.cc,tests/submitperftest.py.in, tests/urllib2_file.py: Fix display of upper bound on number of results of search (was showing lower bound). Fix display of closing <repetition> tags. Add script to submit the results of a performance test to a central server. Tue Apr 01 11:19:58 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc,backends/flint/flint_table.h: Gently rearrange where and how we check for overlong keys. * tests/api_wrdb.cc: Extend termtoolong1 to test that the "term too long" exception messages contain the length correctly. Tue Apr 01 06:06:43 GMT 2008 Olly Betts <olly@survex.com> * AUTHORS,HACKING,INSTALL,README,debian/control.in,debian/copyright, docs/install.html,xapian-core.spec.in: Standardise URLs on our website to be xapian.org not www.xapian.org. Tue Apr 01 03:44:32 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_record.cc,backends/flint/flint_values.cc, common/,docs/quickstartexpand.cc.html,docs/quickstartindex.cc.html, docs/quickstartsearch.cc.html,matcher/: Update the FSF address for the small number of files which still have the old one. Remove the "START_LICENCE" and "END_LICENCE" markers from the small number of files which still have them. Tue Apr 01 02:23:27 GMT 2008 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remoteprotocol.h, common/remoteserver.h,common/serialise.h,docs/remote_protocol.html, net/remoteserver.cc,net/serialise.cc: Update remote protocol to a new major version (31) and strip out all the gunk that's only there to keep it compatible with older version 30 clients. Tue Apr 01 00:12:38 GMT 2008 Olly Betts <olly@survex.com> * api/ompostlistiterator.cc,include/xapian/postingiterator.h: Remove the method Xapian::PostingIterator::get_weight() which has been commented out almost forever. Weight-handling isn't really appropriate here. * api/ompostlistiterator.cc: Remove "\todo" since it's already done! Mon Mar 31 13:24:11 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/: Remove unnecessary default dtors. Mon Mar 31 13:21:55 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Document to prefer "new SomeClass" to "new SomeClass()". * api/omdocument.cc,api/omenquire.cc,backends/database.cc, backends/flint/flint_database.cc, backends/inmemory/inmemory_database.cc,matcher/multimatch.cc, matcher/queryoptimiser.cc,net/remoteserver.cc: Fix instances of the latter form to use the former form. Mon Mar 31 12:11:09 GMT 2008 Olly Betts <olly@survex.com> * tests/: Update svn:ignore property for perftest changes and remove quartz-related entries. * tests/perftest.cc: Add missing '#include "utils.h"'. Mon Mar 31 09:22:35 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * Makefile.am,tests/,tests/harness/,tests/perftest.cc,tests/perftest.h, tests/perftest_matchdecider.cc: Rework the test collation mechanism slightly to allow it to be used by other testsuites: rename collate-apitest to collate-test, and add a parameter to it naming the output file. Add new "testrunner.h" and "testrunner.cc" which are an equivalent to the old mechanism in apitest for running tests with various properties; the properties for each backend are now held as member variables of a TestRunner class, and testsuites (eg, apitest) subclass this and implement a "run()" method, which will be called for each backend with the appropriate properties set. Move definition of DEFINE_TESTCASE macro into testsuite.h. Add new testsuite "perftest", which is intended to contain potentially time consuming performance tests, logging the results to an XML file for later analysis. Add a simple test of the ValueSetMatchDeciders to this testsuite. "perftest" isn't run by make check, because it is likely to take some hours to complete in future - instead, there's a new top-level target "check-perf" which builds and runs perftest. Mon Mar 31 08:13:21 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Recommend using the "svn-ci" script. Update the date command which produces the correct format timestamp for ChangeLog entries. Mon Mar 31 08:00:58 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_btreeutil.h: Fix some out-of-date comments. * configure.ac: Use AC_CHECK_SIZEOF to define SIZEOF_INT and SIZEOF_LONG. * backends/flint/flint_types.h: Use SIZEOF_INT and SIZEOF_LONG to determine the type of uint4 rather than always using unsigned long (which is 64 bits on most 64 bit Unix platforms). Drop int4 for the time being, as we don't actually use it. Mon Mar 31 06:34:20 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Refactor BitWriter::encode() to have a single call to write_bits(), and don't include it inline in the class as it's really a bit big to inline (and write_bits() should now be inlined into it, or at least tail-called). Mon Mar 31 05:33:02 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Reorder header includes to match our new standard more closely. Indent class definitions to match our coding standards. Mon Mar 31 05:31:59 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.h: Remove unnecessary default dtor. Mon Mar 31 04:55:59 GMT 2008 Olly Betts <olly@survex.com> * backends/remote/net_document.cc,backends/remote/net_document.h: Rename "doc" member to "data" since it actually holds the document data. Mon Mar 31 03:25:21 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_document.cc,backends/flint/flint_document.h: Remove unnecessary default dtor. Mon Mar 31 02:53:48 GMT 2008 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_alltermslist.cc, backends/inmemory/inmemory_alltermslist.h: Remove unnecessary default dtor. * backends/inmemory/inmemory_alltermslist.h: Fix bogus cut-and-pasted doxygen comment for ctor. Mon Mar 31 02:33:15 GMT 2008 Olly Betts <olly@survex.com> * common/inmemory_positionlist.h: Remove unnecessary default destructor. Fix comment type ("inemory" -> "inmemory"). Sun Mar 30 12:47:50 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Drop explicit dtor for SimpleStopper which does nothing. Sun Mar 30 12:12:12 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/valuesetmatchdecider.h: Don't bother checking if an element is present before calling std::set::erase(). Don't bother defining an explicit dtor which does nothing. Sat Mar 29 00:46:13 GMT 2008 Olly Betts <olly@survex.com> * tests/api_db.cc: Test matchfunctor bounds in combination with collapsing and percentage cutoff. Also, range check get_matches_estimated rather than checking it's exactly the value currently returned. Fri Mar 28 21:37:27 GMT 2008 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Test ValueSetMatchDecider::remove_value() for a value which isn't in the set. Test that removing a value doesn't affect other values in the set. Fri Mar 28 21:03:30 GMT 2008 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Avoid needless call to pl->get_termfreq_est() if we're using a matchdecider or matchspy. Fri Mar 28 10:28:51 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: Set matches_lower_bound to 0 before starting the match process if there is a match decider or spy; previously, the lower_bound wasn't being reduced in the presence of a match decider unless all the potential results were retrieved. * tests/api_db.cc: Extend matchfunctor1 to check the bounds and estimates returned; includes a regression test for the above. Fri Mar 28 09:09:54 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/api_nodb.cc: Extend tests to check ValueSetMatchDecider remove_value and also ValueSetMatchDeciders holding more than one value. Fri Mar 28 02:41:45 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: Update version info to match 1.0.6. Fri Mar 28 00:12:48 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omenquire.cc,include/xapian/enquire.h: Merge two of the get_mset() variants into one, with default arguments, which has exactly the same effect. We added the two variants to preserve ABI compatibility, and now we're branched for 1.1 this is no longer necessary. Thu Mar 27 23:54:39 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/valuesetmatchdecider.cc,include/xapian.h, include/xapian/valuesetmatchdecider.h: Add ValueSetMatchDecider, which is a matchdecider which is intended to be passed a set of values to look for in documents, and selects documents based on the presence of those values. * tests/api_db.cc,tests/api_nodb.cc: Add tests for the ValueSetMatchDecider - both for calling it standalone, and when passed to get_mset(). * api/Makefile.mk,include/Makefile.mk: Tell build system about new files. Thu Mar 27 06:18:52 GMT 2008 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Initialise term_pos in previous change to fix valgrind-detected error. Thu Mar 27 06:13:01 GMT 2008 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Optimise Query(OP_VALUE_GE, <n>, "") to Query::MatchAll. * tests/api_nodb.cc: Add testcase for this case. Thu Mar 27 04:34:40 GMT 2008 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_document.cc, backends/inmemory/inmemory_document.h: Rename "doc" member to "data" since it actually holds the document data. * backends/inmemory/inmemory_database.cc: Remove unused <stdio.h>. Include "inmemory_database.h" right after <config.h>. Wed Mar 26 23:44:03 GMT 2008 Olly Betts <olly@survex.com> * api/omenquire.cc: Throw InvalidArgumentError upon RSet::add_document(0). * tests/api_nodb.cc: Add regression test for this. * matcher/rset.cc: Add assertion that we don't see a zero docid when calculating rtermfreqs. * api/omdocument.cc: Use RETURN not return so debug logging sees the return value. * backends/inmemory/inmemory_database.cc, backends/remote/remote-database.cc: Change exception throwing for a zero docid to assertions since zero docids should be checked for before we get to here. Wed Mar 26 06:17:51 GMT 2008 Olly Betts <olly@survex.com> * common/postlist.h: Decree that PostList::check() must be passed a docid which actually exists in the database (which is always true currently and seems unlikely to be an onerous requirement). * matcher/valuegepostlist.cc,matcher/valuerangepostlist.cc: Take advantage of this new precondition to simplify the code. * matcher/valuegepostlist.cc,matcher/valuerangepostlist.cc: Move the corresponding header first as our header inclusion order guidelines recommend. Thu Mar 13 01:58:18 GMT 2008 Olly Betts <olly@survex.com> * common/stats.h,include/xapian/enquire.h,matcher/localmatch.cc, matcher/weight.cc: Untangle Xapian::Weight to just initialise internal to 0 in its ctor, which is an ABI change. Resolves a load of FIXME:1.1 comments. Thu Mar 13 00:50:06 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc: Remove workaround for newlib which we don't seem to need elsewhere and was marked "FIXME:1.1". Thu Mar 13 00:49:27 GMT 2008 Olly Betts <olly@survex.com> * docs/deprecation.rst: Fix footnote ids after deprecation updates. Wed Mar 12 10:08:37 GMT 2008 Olly Betts <olly@survex.com> * tests/stemtest.cc: Remove support for OM_STEMTEST_SKIP_RANDOM, OM_STEMTEST_LANGUAGES, and OM_STEMTEST_SEED. Wed Mar 12 08:30:48 GMT 2008 Olly Betts <olly@survex.com> * docs/deprecation.rst: Update for bindings removals for 1.1.0. Wed Mar 12 06:59:25 GMT 2008 Olly Betts <olly@survex.com> * docs/deprecation.rst: Update for Omega removals for 1.1.0. Wed Mar 12 05:11:53 GMT 2008 Olly Betts <olly@survex.com> * api/omenquire.cc,api/omquery.cc,api/version.cc,docs/, generate-exceptions.in,include/xapian.h,include/xapian/enquire.h, include/xapian/query.h,tests/api_db.cc,tests/api_wrdb.cc, tests/apitest.cc,tests/harness/,tests/internaltest.cc: Remove all xapian-core features marked for removal in 1.1.0. Wed Mar 12 03:28:18 GMT 2008 Olly Betts <olly@survex.com> * HACKING,INSTALL,Makefile.am,backends/Makefile.mk, backends/dir_contents,backends/quartz/,bin/,configure.ac,docs/, tests/,tests/harness/,tests/quartztest.cc,tests/testdata/, xapian-core.spec.in: Remove the quartz backend. Wed Mar 12 02:14:33 GMT 2008 Olly Betts <olly@survex.com> * m4/xapian.m4: Improve wording of the error message when we can't find xapian-config. Wed Mar 12 02:13:32 GMT 2008 Olly Betts <olly@survex.com> * INSTALL: Minor wording tweak. Wed Mar 12 01:44:42 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Initialise PostlistCursor member firstdid to avoid using it uninitialised when merging user metadata and the same key is present in more than one source database. If the same user metadata key is present in more than one source database, copy an arbitrary tag value. Wed Mar 12 00:34:04 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Fix potential SEGV (which I think can occur when compacting database(s) with user metadata but no postings). Fri Mar 07 09:46:38 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/serialise.h: Add missing std:: before string. Fri Mar 07 00:47:06 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc: Tidy up a line which was wider than 80 columns, and improve the error message returned when a database replica doesn't contain exactly one subdatabase. Fri Mar 07 00:34:28 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/serialise.h,net/remoteconnection.cc,net/serialise.cc: Change encode_length() to a template, to allow the full range of off_t values to be encoded. Also, adjust RemoteConnection::get_message_chunked() to allow the message length to be a 64 bit quantity. Also, make a couple of type conversions to size_t explicit, to remove some warnings on windows. Thu Mar 06 01:57:04 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog. Thu Mar 06 01:44:07 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Fix terminology - "user metadata" not "user metainfo". Fix bug - need to invoke is_user_metadata rather than using the function pointer as the if condition! Thu Mar 06 01:41:09 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Fix terminology - "user metadata" not "user metainfo". Wed Mar 05 21:17:02 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: Add some casts (in the windows-specific code) to fix some warnings from MSVC. Wed Mar 05 19:01:35 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Update to 1.0 branch point. Wed Mar 05 19:00:07 GMT 2008 Olly Betts <olly@survex.com> * common/msvc_dirent.cc: Tweak to bring comment back into 80 columns. Wed Mar 05 16:35:33 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/msvc_dirent.cc: Fix sense of check for ENOENT after calling _findnext() - it should be checked for when an error code is returned, not when the call succeeds. * tests/api_replicate.cc: Fix tiny capitalisation typo. Wed Mar 05 16:29:47 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/fileutils.cc: Check case where a backslash isn't found in calc_dirname, and don't replace slash with backslash in this case. Wed Mar 05 09:30:42 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-check.cc: Add understanding of user metainfo keys in the postlist table. Allow the METAINFO key to be absent if the table only contains user metainfo keys. Wed Mar 05 09:01:48 GMT 2008 Olly Betts <olly@survex.com> * common/msvc_dirent.cc,common/msvc_dirent.h: Move licence boilerplate up front where we have it in every other file. Move header guards to the usual location. Detail the fix we've made for not setting errno to ENOENT erroneously. Push the setting of orig_errno down a bit. Wed Mar 05 04:57:44 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: -pedantic isn't actually very useful so just drop it rather than testing for known problem cases and risking unknown problem cases. Wed Mar 05 03:10:24 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * matcher/valuegepostlist.cc,matcher/valuegepostlist.h: Implement skip_to method for ValueGePostList. Previously, the ValueRangePostList::skip_to method was used, which would return incorrect documents. Wed Mar 05 02:54:27 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/fileutils.cc: Search for \ instead of / when looking for a backslash! Wed Mar 05 02:51:49 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/fileutils.cc: Fix incorrect docstring. Tue Mar 04 18:48:25 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/msvc_dirent.cc: Fix readdir() so that it doesn't change errno to ENOENT on reaching the end of the directory, as specified by POSIX. Tue Mar 04 17:54:56 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: Fix for a problem with windows. Because we're using mixed styles of windows file IO, if a file descriptor which actually corresponds to a file was supplied to remote connection, the file pointer wasn't being adjusted after each call to ReadFile or WriteFile. This manifested in the test case for replication, resulting in a (corrupt) 2048 byte changeset file being written (instead of a considerably longer changeset). To work around this, we explicitly move the file pointer ourselves. Tue Mar 04 17:41:38 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/api_replicate.cc: Close the replica before removing the temporary directory, to allow the rmdir to work on windows. Thu Feb 28 16:41:34 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/remoteconnection.h: Include safeunistd.h, to get correct typedef for off_t on windows - previously, the size of the RemoteConnection class was varying depending on the order in which header files were included, resulting in memory corruption. * net/remoteserver.cc: Remove incorrect comment about what the problem was. Sat Feb 23 23:40:56 GMT 2008 Olly Betts <olly@survex.com> * xapian-config.in: Use globbing rather than iterating over the output of ls. Sat Feb 23 23:33:39 GMT 2008 Olly Betts <olly@survex.com> * xapian-config.in: Fix not to repeated headers in --swigheaders in a non-VPATH build. Fri Feb 22 17:20:19 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Add note about preferring std::string(). Fri Feb 22 17:05:11 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * net/remoteserver.cc: Use std::string() instead of std::string(""), since it's a bit cleaner. Fri Feb 22 16:50:32 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * net/remoteserver.cc: Add workaround for bug in MSVC 2005. Fri Feb 22 14:39:03 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_database.cc,backends/flint/flint_table.cc, backends/quartz/btree.cc: Use msvc_posix_rename() instead of rename() when renaming base files, so that an error isn't returned if the destination already exists. Thu Feb 21 17:24:14 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/msvc_dirent.cc: Include msvc_dirent.h instead of dirent.h Thu Feb 21 17:09:01 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/database.h: ReplicationInfo is a struct, not a class, so forward declare it as such. Thu Feb 21 12:16:23 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/Makefile.mk: Add common/safedirent.h to distribution. Thu Feb 21 00:25:33 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * xapian-config.in: Add swigheaders option, listing the header files used by swig, so that we can add the appropriate dependencies to the makefiles in the bindings. Wed Feb 20 20:59:55 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * include/xapian/replication.h: Add a ReplicationInfo structure, and use it to return information about what has been done when the replication methods are called. * api/replication.cc,backends/database.cc, backends/flint/flint_database.cc,backends/flint/flint_database.h, common/database.h,common/replicatetcpclient.h, common/replicatetcpserver.h,net/replicatetcpclient.cc, net/replicatetcpserver.cc: Populate a ReplicationInfo structure when updating a replica, or sending changes to a file descriptor. Also, fix a bug with reading the replication parameters from a file, which caused a full database copy to be performed on every replication request on a newly opened DatabaseReplica object. * bin/xapian-replicate.cc: When doing verbose logging, display the number of database copies and changesets applied, and whether the live database has been updated. * tests/api_replicate.cc: Check that the information returned by the replication methods is as is should be, and add a regression test for always copying the database. Wed Feb 20 11:20:55 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/: Improve reporting of failures to obtain lock due to unexpected errors - the error messages included in the exceptions raised are now more verbose in many situations. Also, if the lock can't be obtained when a database is being created, report the lock failure, not a DatabaseOpeningError - it's more useful to know that the lock attempt failed than that the database wasn't present before the attempt to create it. Tue Feb 19 11:58:22 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-compact.cc: Fix compaction for database which contain user metadata keys. Sat Feb 16 19:53:44 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Fix incorrect example in doccomment. Sat Feb 16 14:46:08 GMT 2008 Olly Betts <olly@survex.com> * AUTHORS: Add Matthew Somerville for Search::Xapian patches. Tue Feb 12 16:34:52 GMT 2008 Olly Betts <olly@survex.com> * docs/quickstart.html: Remove information covered by INSTALL since there's no good reason to repeat it and two copies just risks one getting out of date (as has happened here!) Tue Feb 12 16:27:30 GMT 2008 Olly Betts <olly@survex.com> * AUTHORS: Add David Spencer for reporting bug#237. Tue Feb 12 16:25:00 GMT 2008 Olly Betts <olly@survex.com> * docs/quickstart.html: Fix very out of date reference to MSet::items (bug#237). Tue Feb 12 12:40:47 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-compact.cc: Fix incorrect parameters passed to FlintTable constructor. Sat Feb 09 11:56:40 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/socket_utils.h,net/remoteconnection.cc: More fixes for windows. Sat Feb 09 11:31:30 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/msvc_dirent.cc,common/msvc_dirent.h: Hopefully this is a better implementation of dirent functions. Sat Feb 09 10:59:50 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/: Add implementation of dirent.h for use with MSVC; this is copied from the mingw implementation, which has been placed in the public domain. Add "safedirent.h" header for easy inclusion of this implementation. Sat Feb 09 10:43:49 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/api_replicate.cc: Use _putenv instead of setenv on windows. Sat Feb 09 09:56:08 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/socket_utils.cc,common/socket_utils.h: More fixes for windows. Fri Feb 08 16:40:56 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/Makefile.mk,common/socket_utils.cc,common/socket_utils.h, net/remoteconnection.cc,net/tcpclient.cc: Move windows-specific socket handling code from remoteconnection.cc into a separate file, provides the inline close_fd_or_socket for unix in the header file. Use this in tcpclient.cc instead of close to close sockets correctly on windows. Fri Feb 08 16:40:29 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/fileutils.cc,common/replicatetcpclient.h: Correct typos in windows only sections of code. Fri Feb 08 15:29:28 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: Fixes for windows - always include "safesysselect.h", and include <io.h> on windows. Also, remove definition of an unused variable (probably due to a copy and paste error). Wed Feb 06 13:45:59 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-replicate.cc: Include "safeunistd.h" - needed for sleep, if nothing else. Tue Feb 05 18:10:23 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Set -pedantic in CXXFLAGS not AM_CXXFLAGS when testing for problem with it - AM_CXXFLAGS is an automake thing, and isn't used by the autoconf test. Tue Feb 05 17:50:55 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Cleaner patch for the test of compilation with -pedantic - cache the result using AC_CACHE_VAL(). Tue Feb 05 17:32:11 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Check whether compilation of a cast of INFINITY to a double, and comparison with HUGE_VAL, causes a problem if -pedantic is specified (which it does on at least some Mac OSX versions). If so, don't use the -pedantic compiler flag. Tue Feb 05 09:24:36 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_cursor.cc,backends/flint/flint_database.cc, backends/quartz/bcursor.cc: More fixes for debug logging - rename the "hex_encode" functions used for logging to "hex_display_encode" to avoid name clash. Fix return type declared in flint_database.cc's DEBUGCALL macro. Tue Feb 05 09:18:29 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc: Fix for logging compiles: don't check return value of a void function. Tue Feb 05 09:14:34 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc: Reopen database after replication, to fix bug with applying a chain of changesets. * backends/flint/flint_database.cc: Don't use O_TRUNC (or O_CREAT) when modifying the DB files! * tests/api_replicate.cc: Don't call function from inside TEST_EQUAL, because that's a macro and the function will be double-evaluated. Tue Feb 05 02:07:08 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,backends/database.cc,backends/flint/, common/database.h: Many fixes to the replication code. Now parses replication changesets correctly and applies them to the databases. Also, correct one bug with the generation of changesets. * docs/replication.rst,docs/replication_protocol.rst: Update. * tests/api_replicate.cc: Add test of two changesets being applied at once. Sun Feb 03 13:31:10 GMT 2008 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Backport atomic base update fix from flint. Sat Feb 02 22:47:56 GMT 2008 Olly Betts <olly@survex.com> * AUTHORS: Add Thomas Viehmann. Sat Feb 02 03:39:33 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Update the base files atomically to avoid problems with reading processes finding partially written ones. Sat Feb 02 03:38:00 GMT 2008 Olly Betts <olly@survex.com> * api/replication.cc: Don't leak the RemoteConnection object. Fri Feb 01 23:15:38 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_btreebase.cc,backends/flint/flint_btreebase.h: Fix assorted comment errors. Fri Feb 01 20:20:23 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc: Fix comment. Fri Feb 01 20:13:23 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-replicate.cc: Document the default interval in --help. Fri Feb 01 20:05:08 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-replicate.cc: Remove the "return 0;" I added at the end of the function - I didn't realise this, but in C++ it's not necessary to have a return at the end of main(). Tidier without it, so it's gone. Fri Feb 01 19:44:05 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Add Unicode::toupper() to complement Unicode::tolower(). * tests/api_unicode.cc: Add caseconvert1 testcase to test Unicode::tolower() and Unicode::toupper(). Fri Feb 01 19:36:47 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-replicate.cc: Add one-shot and verbose options to replication client. Fri Feb 01 17:52:36 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,backends/flint/flint_database.cc, include/xapian/replication.h,net/replicatetcpclient.cc, tests/api_replicate.cc: Keep the RemoteConnection in the DatabaseReplica object, so that any data which is read ahead doesn't get lost. If an error occurs opening the database on the server, send a FAIL message over the connection, rather than just shutting it. Fri Feb 01 16:51:05 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_table.cc: Fix a database corruption bug: see bugzilla entry #232. Was caused by lazy tables not being created at the right revision. * tests/api_wrdb.cc: Regression test for bug. Fri Feb 01 15:10:23 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,backends/database.cc, backends/flint/flint_database.cc,backends/flint/flint_database.h, common/database.h,docs/replication_protocol.rst, tests/api_replicate.cc: Add the UUID of the new database to the DB_HEADER message, and move the code which compares UUIDs out of flint_database.cc and into replication.cc. Fix testcase which missed incorrect comparison of UUIDs. Store the UUID of the live database which has been replicated in the config file - this isn't an ideal place, but will do until the UUID is stored in the iamflint file, and we can parse it from there. Remove debug printfs. Fri Feb 01 15:09:12 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * bin/quartzdump.cc: Rename hex_encode() to hex_display_encode() to avoid collision. (The hex encoding produced by this doesn't encode spaces, so isn't quite the same as the functions just added to utils.cc). Fri Feb 01 15:08:06 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/utils.cc,common/utils.h: Add utility functions to hex-encode and decode a string, so we can store UUIDs in the config file in replicated DB directories. Fri Feb 01 12:49:34 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/harness/backendmanager_multi.cc: Make the multi backendmanager use relative paths in its stub databases. Fri Feb 01 12:17:38 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/Makefile.mk: Add replicationprotocol.h to distribution. Fri Feb 01 12:10:19 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/database.cc: Stub databases used to assume that any relative paths were relative to the current working directory. They now assume that relative paths are relative to the directory holding the stub database file. Also, lines which begin with a '#' character are ignored, so we can place comments in the stub database files. Also, recognise a new database type: a "stub directory", which is a directory containing a stub database file named "XAPIANDB". * api/replication.cc: Replica databases are now created as directories containing a "XAPIANDB" stub file, so all the workings are internal. * common/Makefile.mk,common/fileutils.cc,common/fileutils.h: Add new set of utilities to manipulate path names - extracting directory names, and joining paths. * tests/api_replicate.cc: Enable test properly, now that it passes. Test needs to be expanded to check that the database copy succeeded better. Fri Feb 01 09:47:00 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_database.cc: Fix warning about unused variable. Fri Feb 01 03:49:56 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,backends/database.cc, backends/flint/flint_database.cc,backends/flint/flint_database.h, common/database.h,common/replicationprotocol.h: Move replication protocol definition into a separate file. Add (virtual) support methods for applying changesets to Database::Internal. Sort out atomic swapping of old database for new after a database copy. Fri Feb 01 03:49:09 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/quartztest.cc: Rename removedir to removedir_recursive() to avoid conflict with my new function. Fri Feb 01 03:09:45 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * docs/replication_protocol.rst: Specify (though not in much detail) what we mean by "packed" strings and integers. Fri Feb 01 03:09:06 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * include/xapian/replication.h: Be more precise in documentation comment. Fri Feb 01 03:07:29 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: Remove debug printfs() Fri Feb 01 03:03:59 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/utils.cc,common/utils.h: Add overloaded rmdir() which works directly on C++ strings. Add removedir() method which removes a directory and its contents (as long as the directory only contains files, not subdirectories). Fri Feb 01 01:51:22 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/remoteconnection.h,net/remoteconnection.cc: Add "sniff_next_message_type" method to RemoteConnection, and replace the hardcoded "4096" size of the chunks used for the remote connection confersation with a #define. Fix file mode when creating a file with receive_file(), and bugs with writing too much data to the file. Thu Jan 31 03:41:42 GMT 2008 Olly Betts <olly@survex.com> * bin/Makefile.mk,bin/xapian-replicate-server.cc, common/replicatetcpserver.h: Add xapian-replicate-server server program. Thu Jan 31 02:46:38 GMT 2008 Olly Betts <olly@survex.com> * bin/Makefile.mk,bin/xapian-replicate.cc,common/replicatetcpclient.h, net/replicatetcpclient.cc,net/replicatetcpserver.cc: Add xapian-replicate client program. Wed Jan 30 19:13:08 GMT 2008 Olly Betts <olly@survex.com> * include/xapian/replication.h: Correct name of parameter in doxygen comment. Wed Jan 30 15:58:44 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,backends/database.cc, backends/flint/flint_database.cc,backends/flint/flint_database.h, common/database.h: Hook up the apply_changeset function from the API through to flint. (The flint layer currently doesn't work, though.) Also, implement the get_uuid() function for flint, by using the mtime of the iamflint file. * tests/api_replicate.cc: Finish implementation of the "replicate" function, and add a test of a replication of a database using this. The return value test is commented out for now, because the flint layer doesn't work, but this puts the machinery needed to test this in place. Wed Jan 30 15:10:37 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_database.cc: Implement sending of the whole database copy. Wed Jan 30 14:45:01 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * tests/,tests/harness/: Add new testcase file (api_replicate.cc) to hold tests of the replication functionality. Currently has one partially written test. Also, add new function to apitest.h: get_named_writable_database_path(), which allows the path to a writable database to be obtained; throws an exception for those backend types for which that isn't meaningful. Add a new variable for use in the conditions for tests: "replicas", which should be set to true for all backends which support replication. Wed Jan 30 14:42:26 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: Fix the calls to fcntl() in send_message() and send_file() to use fdout instead of fdin. This has probably been stopping timeouts working very well. Tue Jan 29 15:05:09 GMT 2008 Olly Betts <olly@survex.com> * common/remoteconnection.h,net/remoteconnection.cc: Add receive_file() method. Tue Jan 29 14:37:36 GMT 2008 Olly Betts <olly@survex.com> * common/remoteconnection.h,net/remoteconnection.cc: Add the ability to read a message in chunks. Tue Jan 29 10:08:02 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * include/xapian/replication.h: Expand documentation comment. * api/replication.cc,backends/database.cc, backends/flint/flint_database.cc,backends/flint/flint_database.h, common/database.h: Add a write_changesets_to_fd() function to Database::Internal, with a default implementation to raise an UnimplementedError. Hook DatabaseMaster::write_changesets_to_fd() up to this, and implement it for flint databases (the code to send a copy of a database still needs work). Tue Jan 29 10:01:39 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * docs/replication_protocol.rst: Document that a FAIL message ends the conversation. Tue Jan 29 09:59:01 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * docs/replication_protocol.rst: Document the protocol used to transfer the updates. Tue Jan 29 09:57:14 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: Fix typo. Tue Jan 29 09:48:15 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: For win32, use msvc_posix_open() to open the file descriptor, so other things can delete the file while the sending is in progress. Tue Jan 29 01:23:19 GMT 2008 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Add an assertion. Tue Jan 29 01:14:07 GMT 2008 Olly Betts <olly@survex.com> * common/remoteconnection.h,net/remoteconnection.cc: Add send_file() method to RemoteConnection to allow the contents of a file to be sent as a message. Tue Jan 29 00:28:18 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,include/xapian/replication.h: Add methods to allow parameters to be stored associated with a DatabaseReplica. Tue Jan 29 00:16:42 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_database.cc,backends/flint/flint_database.h: Change write_changesets_to_fd() to take the revision information as a string - it now includes a UUID. Add get_uuid() method to FlintDatabase - currently always returns "FIXME", but will eventually return an ID for the database. Mon Jan 28 19:07:20 GMT 2008 Olly Betts <olly@survex.com> * common/Makefile.mk,common/replicatetcpclient.h,net/Makefile.mk, net/replicatetcpclient.cc,net/replicatetcpserver.cc: Add ReplicateTcpClient class implementing a replication client over TCP/IP. Mon Jan 28 17:13:09 GMT 2008 Olly Betts <olly@survex.com> * common/Makefile.mk,common/replicatetcpserver.h,net/Makefile.mk, net/replicatetcpserver.cc: Add ReplicateTcpServer class implementing a replication server over TCP/IP. Mon Jan 28 17:01:02 GMT 2008 Olly Betts <olly@survex.com> * common/remoteconnection.h,net/remoteconnection.cc: Minor tweaks to allow unidirectional connections. Mon Jan 28 16:26:22 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc: Fix overly hasty commit - add get_description() method for internal class, and use that instead of trying to access a private member. Mon Jan 28 16:22:18 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,common/output.h,include/xapian/replication.h: Add get_description() methods to DatabaseMaster and DatabaseReplica, and add XAPIAN_OUTPUT_FUNCTION methods for them so that they can be displayed in debugging methods. Should fix build with logging enabled. Mon Jan 28 11:52:24 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,include/xapian/replication.h: Add DatabaseReplica::close() method, to allow the write lock to be released easily. Mon Jan 28 11:21:36 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,include/xapian/replication.h: Change DatabaseReplica into a PIMPL style class, now that it has more than one member. Add visibility annotations to DatabaseMaster and DatabaseReplica. Mon Jan 28 10:44:51 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,include/xapian/replication.h: Add code to swap stub database files over atomically, and to track the current real database path. Mon Jan 28 04:52:12 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/replication.cc,backends/database.cc, backends/flint/flint_changesetapplier.h, backends/flint/flint_database.cc,backends/flint/flint_database.h, common/database.h: Implement opening of the database underlying DatabaseReplica objects, and hook up the get_revision_info() method for it. Mon Jan 28 04:17:21 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * common/utils.cc,common/utils.h: Add "dir_exists" function, to check for presence of a directory at a given path. Mon Jan 28 03:48:46 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_table.cc: Change a "+= 1" to ++ - this isn't Python! Mon Jan 28 03:08:45 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/Makefile.mk,api/replication.cc,include/Makefile.mk, include/xapian/replication.h: Add interface to replication functionality. (Currently just unimplemented stubs.) Mon Jan 28 02:01:02 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_btreebase.h: Add a missing std:: Mon Jan 28 01:59:55 GMT 2008 Olly Betts <olly@survex.com> * common/remotetcpserver.h: Fix typo in doxygen comment: `@port' -> `@param port'. Update comment - socket is no longer closed by handle_one_connection(). Mon Jan 28 01:28:00 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/: Add support for recording the changes made to a database in "changeset" files. These are currently not produced by default - for now, the "XAPIAN_MAX_CHANGESETS" environment variable can be set to control production of them. As part of these changes, the FlintTable constructor now takes the tablename, so that this can be stored in changesets produced from the table. * bin/xapian-check.cc,bin/xapian-compact.cc,bin/xapian-inspect.cc: Update to match changes in FlintTable constructor. * docs/Makefile.am,docs/replication.rst,docs/replication_protocol.rst: Add basic documentation of the replication protocol (though this isn't yet definitive, since the protocol isn't finished!) Sun Jan 27 21:10:24 GMT 2008 Olly Betts <olly@survex.com> * net/remotetcpserver.cc,net/tcpserver.cc: The job of closing the socket connected to the client more naturally belongs to the framework TcpServer class. Sun Jan 27 20:32:55 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Call flint_io_sync() in commit() rather than flush_db(), as that gives more time for written blocks to get written to disk before we block on waiting for them to be. This should tend to speed up WritableDatabase::flush() in I/O bound situations. Sun Jan 27 19:35:08 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_spelling.h,backends/flint/flint_synonym.h: Merge changes to the spelling and synonym tables just before calling FlintTable::flush_db(), rather than just before calling FlintTable::commit(), so that they don't get forgotten. Sun Jan 27 14:29:30 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_database.cc,backends/flint/flint_table.cc, backends/flint/flint_table.h: Separate step which flushes modified blocks from cursors to the DB file from the rest of commit. This will allow me to insert the changeset writing code between the two steps, so it can assume that the DB file is fully up-to-date. Also remove an out-of-date FIXME about trying to avoid updating the value and position tables if they're not used: these tables are now lazily created anyway. Sun Jan 27 12:53:51 GMT 2008 Olly Betts <olly@survex.com> * net/tcpserver.cc: Change explicit references to xapian-tcpsrv. Sun Jan 27 12:26:24 GMT 2008 Olly Betts <olly@survex.com> * common/tcpclient.h,common/tcpserver.h,net/: Make use of TCP_NODELAY optional at the class level. No user-visible changes. Sat Jan 26 17:18:09 GMT 2008 Olly Betts <olly@survex.com> * backends/dbfactory_remote.cc,common/Makefile.mk, common/remotetcpclient.h,common/tcpclient.h,net/Makefile.mk, net/remotetcpclient.cc,net/tcpclient.cc: Split out the "open a connection" part of TcpClient and rename the rest to RemoteTcpClient. Sat Jan 26 14:01:34 GMT 2008 Olly Betts <olly@survex.com> * bin/xapian-tcpsrv.cc,common/Makefile.mk,common/remotetcpserver.h, common/tcpserver.h,net/Makefile.mk,net/remotetcpserver.cc, net/tcpserver.cc: Split TcpServer into a generic TCP server class (TcpServer) and a subclass of this which implements the remote backend TCP server (RemoteTcpServer). Sat Jan 26 11:06:30 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_btreebase.cc,backends/quartz/btree_util.h, common/utils.h: Move fdcloser into utils.h, since it's generally useful, and it's better not to have multiple copies of it scattered through the code. Fri Jan 25 15:06:30 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_table.cc,backends/quartz/btree.cc: Use msvc_posix_unlink() on windows in sys_unlink_if_exists() functions, so that they will work even if the file is still open. This probably changes no behaviour, because sys_unlink_if_exists() is probably never called in such a situation, but this may not be true in future. Thu Jan 24 13:31:20 GMT 2008 Olly Betts <olly@survex.com> * docs/overview.html: Remove commented-out comment about OP_XOR. Thu Jan 24 13:30:04 GMT 2008 Olly Betts <olly@survex.com> * docs/intro_ir.html: Briefly mention how pure boolean retrieval is supported. Thu Jan 24 13:28:44 GMT 2008 Olly Betts <olly@survex.com> * HACKING: Move "debian/patch" update earlier in the checklist. Thu Jan 24 13:26:45 GMT 2008 Olly Betts <olly@survex.com> * NEWS: Partly update from ChangeLog. Thu Jan 17 22:32:08 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc,backends/flint/flint_lock.h, bin/xapian-check.cc: Apply tweaked version of patch for OS/2 support by Yuri Dario. * AUTHORS: Add Yuri Dario. * PLATFORMS: Mention OS/2. Mon Jan 14 11:03:05 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_table.cc,backends/quartz/btree.cc: Fix more assertions to avoid overflowing int. Mon Jan 14 09:29:51 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_btreeutil.h: Fix assertions to not overflow int. Mon Jan 14 09:23:54 GMT 2008 Olly Betts <olly@survex.com> * backends/quartz/btree_util.h: Fix assertions to not overflow int. Mon Jan 14 08:47:33 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Fix test for gcc being 4.0.x or 4.1.x, so we don't use -Wstrict-overflow with them. Sun Jan 13 23:28:04 GMT 2008 Olly Betts <olly@survex.com> * configure.ac: GCC warning flag overhaul. Stop passing "-Wno-multichar" since any multi-character character literal is bound to be a typo (I believe we were only passing it after misinterpreting its sense!) Pass "-Wformat-security", "-Wconversion", and "-pedantic" for all GCC versions. Add "-Winit-self" and "-Wstrict-overflow=5" for GCC >= 4.2. The latter may be too aggressive, but it's hard to know without trying it more widely. Sun Jan 13 00:25:25 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_check.h,backends/flint/flint_table.h, backends/quartz/btree.h,net/tcpserver.cc,tests/harness/testsuite.cc: Add XAPIAN_NORETURN() annotations to functions and non-virtual methods which don't return. * net/remoteserver.cc: Assign bool variable using a comparison rather than subtraction, so the intent is clearer. Sun Jan 13 00:23:41 GMT 2008 Olly Betts <olly@survex.com> * backends/flint/flint_check.cc: Tweak a comparison so all the constants are on the same side (micro-optimisation). * backends/quartz/btreecheck.cc: Equivalent change for quartz. Sun Jan 13 00:20:08 GMT 2008 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add a couple of testcases I wrote for a reported problem which turned out to work correctly already. Still, they add to our test coverage. Thu Jan 10 02:04:55 GMT 2008 Olly Betts <olly@survex.com> * tests/runtest.in: Cope with "@EXEEXT" extension on test programs. Wed Jan 09 21:57:59 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omquery.cc,api/omqueryinternal.cc,common/remoteprotocol.h, include/xapian/query.h,matcher/queryoptimiser.cc,tests/api_anydb.cc: Add OP_VALUE_LE operator, for symmetry with OP_VALUE_GE. Currently implemented internally using a ValueRangePostList with an empty string as the start of the range. Wed Jan 09 19:36:37 GMT 2008 Richard Boulton <richard@lemurconsulting.com> * api/omquery.cc,api/omqueryinternal.cc,include/xapian/query.h: Add new query operator OP_VALUE_GE, for performing "return documents with a value greater than this" searches. Also add a new constructor to go with this which takes a value number and a single string argument. * matcher/: Add ValueGePostList class, as a subclass of ValueRangePostList. Change internal members of ValueRangePostList to be protected instead of private. Add support for making ValueGePostLists from OP_VALUE_GE queries. * tests/api_anydb.cc: Add valuege1 testcase of OP_VALUE_GE queries. * common/remoteprotocol.h: Bump minor protocol number, due to additional operator being possible in serialised queries. Wed Jan 09 15:18:16 GMT 2008 Olly Betts <olly@survex.com> * PLATFORMS: Update from Debian buildd logs. Wed Jan 09 13:53:34 GMT 2008 Olly Betts <olly@survex.com> * tests/harness/backendmanager_remotetcp.cc: Fix XAPIAN_NORETURN() to wrap a declaration, not the definition, to fix compile failure on mingw "make check". Correct file documentation comment to refer to "remotetcp" not "remoteprog". Mon Jan 07 01:28:14 GMT 2008 Olly Betts <olly@survex.com> * PLATFORMS: Remove reports for 0.8.x as they're too old to be interesting. Separate out 0.9.x reports. Mon Jan 07 00:23:10 GMT 2008 Olly Betts <olly@survex.com> * PLATFORMS: Add Solaris 9 and 10 success reports from James Aylett. Sat Jan 05 19:13:12 GMT 2008 Olly Betts <olly@survex.com> * languages/compiler/generator.c: In generate_call(), if the failure case would just be "if (ret == 0) return 0;" then combine it with the test for the called method returning signal `f'. Sat Jan 05 18:28:31 GMT 2008 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Generate more readable code for the inlined single ASCII character literal string check. Sat Jan 05 18:27:07 GMT 2008 Olly Betts <olly@survex.com> * languages/steminternal.cc: Tweak skip_utf8() to save ~0.5% on stemtest. Sat Jan 05 02:20:57 GMT 2008 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Optimise a single ASCII character literalstring check by inlining the check. This makes stemtest run about 2% faster. Fri Dec 21 21:57:11 GMT 2007 Olly Betts <olly@survex.com> * docs/valueranges.rst: Fix example of using multiple VRPs to come out as a "program listing". Fri Dec 21 15:31:35 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/stringutils.h,common/utils.h: Move declaration of within_DBL_EPSILON back into utils.h to fix builds with --enable-assertions, and since it's not string-related. Fri Dec 21 02:12:49 GMT 2007 Olly Betts <olly@survex.com> * NEWS: Note that one change was "(bug#45)". Bump release date. Thu Dec 20 22:00:41 GMT 2007 Olly Betts <olly@survex.com> * NEWS: Update again for 1.0.5. Thu Dec 20 17:50:26 GMT 2007 Olly Betts <olly@survex.com> * common/stringutils.cc: Add new file I failed to commit. Thu Dec 20 17:17:28 GMT 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add feature tests to ensure that ':' is inserted between prefix and term when it should be. Thu Dec 20 13:58:18 GMT 2007 Olly Betts <olly@survex.com> * common/,queryparser/queryparser.lemony, queryparser/termgenerator_internal.cc,tests/harness/index_utils.cc, tests/harness/unixcmds.cc: Move C_isupper(), C_toupper(), etc from utils.cc/utils.h to stringutils.cc/stringutils.h, since they are string-related. Thu Dec 20 03:00:08 GMT 2007 Olly Betts <olly@survex.com> * HACKING: Remove details of issues with autoconf < 2.57 and automake < 1.5 as these versions are antiques compared to our current requirements of autoconf >= 2.59 and automake >= 1.8.3. Wed Dec 19 03:41:42 GMT 2007 Olly Betts <olly@survex.com> * AUTHORS,NEWS,configure.ac: Update for 1.0.5. Wed Dec 19 03:36:30 GMT 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc: Calling WritableDatabase methods when we don't have exactly one subdatabase now throws InvalidOperationError. * tests/termgentest.cc: Add regression test to check that we now get InvalidOperationError (previously this case gave a segmentation fault). Wed Dec 19 01:39:01 GMT 2007 Olly Betts <olly@survex.com> * PLATFORMS: Update from tinderbox. Tue Dec 18 23:17:37 GMT 2007 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc,backends/flint/flint_lock.h: Move lots of headers included by flint_lock.h to flint_lock.cc since flint_lock.h doesn't need them. Add a "FIXME:1.1:" comment about an odd special-case for newlib. Thu Dec 13 17:34:19 GMT 2007 Olly Betts <olly@survex.com> * tests/stemtest.cc: Mark undocumented OM_STEMTEST_* environmental variables for removal in 1.1. Thu Dec 13 01:27:15 GMT 2007 Olly Betts <olly@survex.com> * HACKING: Need to check RPM packaging with Tim Brody too. Thu Dec 13 01:26:16 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Note that we check for <streambuf> because GCC 2.95 only has <streambuf.h>. Thu Dec 13 01:23:27 GMT 2007 Olly Betts <olly@survex.com> * AUTHORS: Add Ralf Wildenhues for suggesting a way to factor out the boilerplate in multitarget rules. Wed Dec 12 02:02:51 GMT 2007 Olly Betts <olly@survex.com> * NEWS: Update for ChangeLog in preparation for 1.0.5. Wed Dec 12 01:02:29 GMT 2007 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Mark as up to date for Xapian 1.0.5. Minor wording improvements. Sun Dec 09 01:09:19 GMT 2007 Olly Betts <olly@survex.com> * tests/collate-apitest: Check for sources in the build directory first, and only if not there prepend "$srcdir/". The recently added api_generated.cc will be in the build directory when using an SVN checkout. Fri Dec 07 12:37:08 GMT 2007 Olly Betts <olly@survex.com> * docs/sorting.rst,docs/valueranges.rst: State explicitly that Xapian::sortable_serialise() is used to encode values at index time and give an example of how it is called. Thu Dec 06 15:55:31 GMT 2007 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/collate-apitest, tests/generate-api_generated: Add automatically generated tests to ensure that API classes generally have copy ctors and assignment operators, and also generally have a default ctor. Thu Dec 06 15:32:23 GMT 2007 Olly Betts <olly@survex.com> * docs/sorting.rst: Fill in the remaining items in the outline. Thu Dec 06 12:50:46 GMT 2007 Olly Betts <olly@survex.com> * tests/valgrind.supp: Another variant of the zlib suppression. Wed Dec 05 18:31:12 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * Makefile.am,docs/Makefile.am,tests/Makefile.am: Add dir_contents, docs/dir_contents and tests/dir_contents to distribution tarballs. Wed Dec 05 18:28:12 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Really fix handling of version.h. Wed Dec 05 18:20:54 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Handle make pattern substitutions. Wed Dec 05 18:11:31 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Fix special-case handling of include/xapian/version.h. Wed Dec 05 18:01:34 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/multiandpostlist.cc,matcher/multiandpostlist.h: Implement get_wdf() method for MultiAndPostList (by adding together the wdfs of the subpostlists). Not currently used (but will be by SynonymPostList). Wed Dec 05 17:43:24 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Special-case version.h. Wed Dec 05 17:31:47 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Fix errors in previous check-in. Wed Dec 05 17:18:26 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Parse BUILT_SOURCES to determine which files are generated and so need looking for in the build directory - this won't need updating when new files are generated, or existing ones stop being. Wed Dec 05 16:55:26 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * preautoreconf: Fix "make doxygen_docs" with VPATH build: look for generated code files in build dir, and make rule which re-runs preautoreconf change to srcdir first. Wed Dec 05 14:59:52 GMT 2007 Olly Betts <olly@survex.com> * docs/sorting.rst: Document set_sort_by_value() and friends. Wed Dec 05 11:31:27 GMT 2007 Olly Betts <olly@survex.com> * docs/sorting.rst: More work. Tue Dec 04 14:42:57 GMT 2007 Olly Betts <olly@survex.com> * docs/bm25.html: Improve wording. Mon Dec 03 17:11:01 GMT 2007 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/index.html,docs/sorting.rst: Add the start of a topic document on sorting. Mon Dec 03 17:10:04 GMT 2007 Olly Betts <olly@survex.com> * common/stats.h: Fix comment typos. Sun Dec 02 14:36:47 GMT 2007 Olly Betts <olly@survex.com> * docs/synonyms.rst: Minor wording clarification. Sun Dec 02 14:23:37 GMT 2007 Olly Betts <olly@survex.com> * tests/collate-apitest: Fix reversed conditional for deciding which generated headers to update. Sun Dec 02 03:47:44 GMT 2007 Olly Betts <olly@survex.com> * tests/apitest.cc: Include api_all.h instead of the individual generated headers. Sun Dec 02 03:44:56 GMT 2007 Olly Betts <olly@survex.com> * tests/collate-apitest: Add licence and (C). Only update .h files which may have changed. Generate api_all.h which just includes the generated headers corresponding to all the sources processed. Strip spaces from conditions. * tests/Makefile.am: Update for api_all.h. Sun Dec 02 03:21:59 GMT 2007 Olly Betts <olly@survex.com> * matcher/weight.cc: Wrap comment. Sun Dec 02 03:21:25 GMT 2007 Olly Betts <olly@survex.com> * AUTHORS: Thank Petr Ročkai. Sun Dec 02 02:40:58 GMT 2007 Olly Betts <olly@survex.com> * api/omenquire.cc: Oops, this file also part of previous change. Sun Dec 02 02:14:27 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Implement copy ctor and assignment operator for Xapian::Enquire (bug#219). * configure.ac: Note what this means in terms of library versioning. Sat Dec 01 02:54:51 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/: Stop describing get_description() as an "Introspection method", as this doesn't help to explain what it does, and get_description() doesn't actually fall under any of the formal definitions of "introspection" I can find. Sat Dec 01 02:45:16 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/index_utils.h: Overlooked (C) update. Sat Dec 01 02:34:59 GMT 2007 Olly Betts <olly@survex.com> * HACKING,INSTALL,common/,configure.ac,tests/api_anydb.cc, tests/api_db.cc,tests/btreetest.cc,tests/harness/testsuite.cc, tests/harness/testsuite.h,tests/harness/testutils.cc, tests/internaltest.cc: Raise the minimum supported GCC version to 2.95.3 and strip out om_ostringstream which is no longer required. Sat Dec 01 01:10:04 GMT 2007 Olly Betts <olly@survex.com> * tests/api_sorting.cc,tests/api_wrdb.cc: Move sortfunctor2 from api_wrdb.cc to api_sorting.cc. Sat Dec 01 00:43:31 GMT 2007 Olly Betts <olly@survex.com> * tests/valgrind.supp: Add version of the valgrind zlib suppression for 64 bit hosts. Fri Nov 30 18:06:49 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/index_utils.cc: Workaround MSVC which doesn't clear the fail bit when a stream is reopened. Fri Nov 30 00:44:21 GMT 2007 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Make delete_ptr<> a functor rather than a function. This fixes the build with SGI's compiler, and apparently a functor allows more compilers to inline the call. Thu Nov 29 19:24:22 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Fix previous change to work. Thu Nov 29 18:53:09 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Assume we have <sstream> and hardwire HAVE_SSTREAM to be 1. If this assumption proves correct for all platforms we care about, we can rip out the old om_ostringstream code. Thu Nov 29 17:51:11 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/index_utils.cc: Work around MSVC. Wed Nov 28 03:17:06 GMT 2007 Olly Betts <olly@survex.com> * api/sorter.cc: Fix reverse sorting of value strings with different lengths. * tests/api_wrdb.cc: Add test sortfunctor2 to check this actually works. Wed Nov 28 02:27:37 GMT 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc,tests/apitest.cc,tests/apitest.h,tests/harness/: Add get_named_writable_database() and use it to fix spell3 on Microsoft Windows. Tue Nov 27 16:56:12 GMT 2007 Olly Betts <olly@survex.com> * xapian-config.in: Factor out the code to check if we need explicit dependencies into a function. On platforms we know don't need explicit dependencies, --ltlibs now gives the same output as --libs. Tue Nov 27 00:24:18 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am: Remove api_regressions.cc from list - didn't mean to commit that (yet). Tue Nov 27 00:17:40 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am: Use $(collated_apitest_sources) in list of apitest_SOURCES; now a new test file can be added just by adding it to this variable. Mon Nov 26 23:04:00 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am: Add generated apitest .h files to targets of rule which calls collate-apitest, since they're also generated by it. Change the rule to use a .stamp and a .lock file as described in HACKING, to fix parallel builds. Mon Nov 26 18:48:20 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am: Distribute preautoreconf. Mon Nov 26 17:24:15 GMT 2007 Olly Betts <olly@survex.com> * HACKING,configure.ac,docs/Makefile.am: Use pngcrush to reduce the size of PNG files in the doxygen-generated HTML docs. Mon Nov 26 15:39:58 GMT 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Don't package or install various intermediate files which doxygen generates. Mon Nov 26 13:14:19 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_db.cc: Ensure that the database needed for stubdb1 is present, by creating it first (with get_database()). This requires that the database type is flint, but the test doesn't depend on any particular database type anyway, so this doesn't lose us anything. Mon Nov 26 13:08:08 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/collate-apitest: Display the name of the backend which a set of tests is for. Mon Nov 26 09:45:39 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am,tests/collate-apitest: Fix with VPATH builds - pass srcdir as the first argument to collate-apitest, so it can find the sources. Also, use srcdir to form the path to collate-apitest when calling it from the make rule. Mon Nov 26 02:22:08 GMT 2007 Olly Betts <olly@survex.com> * tests/api_db.cc,tests/apitest.cc,tests/apitest.h: Convert the remaining tests to be collated. Mon Nov 26 02:08:18 GMT 2007 Olly Betts <olly@survex.com> * tests/api_db.cc,tests/apitest.cc,tests/apitest.h: Collate the localdb and remotedb tests. Mon Nov 26 01:40:59 GMT 2007 Olly Betts <olly@survex.com> * tests/: Collate the specchar, doclendb, collfreq, allterms, and multivalue tests. Mon Nov 26 01:21:35 GMT 2007 Olly Betts <olly@survex.com> * tests/: Collate the anydb tests. Remove the generated files upon "make clean" in maintainer-mode. Mon Nov 26 00:34:03 GMT 2007 Olly Betts <olly@survex.com> * tests/: Collate the wrdb tests. Mon Nov 26 00:18:18 GMT 2007 Olly Betts <olly@survex.com> * tests/api_posdb.cc,tests/apitest.cc: Add "writable" flag and use it for poslist2 and poslist3. Mon Nov 26 00:13:43 GMT 2007 Olly Betts <olly@survex.com> * tests/: Collate the posdb tests. Mon Nov 26 00:03:29 GMT 2007 Olly Betts <olly@survex.com> * tests/: Collate the nodb tests. Sun Nov 25 23:55:02 GMT 2007 Olly Betts <olly@survex.com> * tests/: Collate the transaction tests. Sun Nov 25 23:49:26 GMT 2007 Olly Betts <olly@survex.com> * tests/: Collate the unicode tests. Sun Nov 25 23:35:40 GMT 2007 Olly Betts <olly@survex.com> * tests/: Make a start on automatically collating test cases. Sun Nov 25 22:41:52 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * api/weightinternal.cc: Initialise all members of Xapian::Weight::Internal, and use initialisers rather than assignment to do so. Fixes an error reported by valgrind. Sun Nov 25 16:48:21 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/weight.cc: Add another note of a FIXME needed for 1.1 Sun Nov 25 16:40:28 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/bm25weight.cc,matcher/tradweight.cc: Add some missing "using namespace std;" lines which are needed due to earlier changes in header file inclusions. Also, remove some unnecessary std:: prefixes. Sun Nov 25 16:35:30 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/stats.h,include/xapian/enquire.h,matcher/localmatch.cc, matcher/weight.cc: Undo ABI change introduced in previous commit with a nasty workaround; we can't initialise Weight::internal reliably, because applications complied with earlier versions of the library may have inlined the constructor of the Weight class. Therefore, we can't delete Weight::internal in the destructor, because we don't know if it was initialised. Instead, we add factory methods for making Weight::Internal objects to Stats, register all the Weight::Internal objects made by these factory methods in a list is the Stats object, and delete them when Stats is deleted (which conveniently happens after the match has finished). Mark all this with FIXME:1.1: so that we can easily convert it to nicer code once we branch for 1.1. Sun Nov 25 15:14:04 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * api/Makefile.mk,api/weightinternal.cc,common/, include/xapian/enquire.h,matcher/: Split Stats and Xapian::Weight::Internal into different classes: Stats remains with the same members (but becomes a concrete class, rather than a typedef of another class), and Xapian::Weight::Internal becomes a simple container for exactly those statistics which are relevant for a weight object (ie, it holds just one termfreq and reltermfreq, rather than a map of all of them). This is dynamically allocated, so Xapian::Weight now owns the object pointed to by its "internal" member, and deletes it in its destructor. Add a new header file "weightinternal.h" so that the full "stats.h" stuff doesn't need to be included for definitions of weighting schemes. Replace #include "stats.h" lines with forward declarations of Stats, where possible. Sun Nov 25 15:09:29 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * api/omenquire.cc: Add missing include of stats.h * matcher/queryoptimiser.cc,net/tcpserver.cc: Remove unnecessary include of stats.h * net/remoteserver.cc: Include missing includes of omassert.h and stats.h, and change an Assert to AssertEq. * matcher/multimatch.cc: Output the contents of the stats object in debugging code. Sun Nov 25 04:22:05 GMT 2007 Olly Betts <olly@survex.com> * common/multimatch.h: No longer need "autoptr.h" here. Sun Nov 25 04:07:04 GMT 2007 Olly Betts <olly@survex.com> * common/multimatch.h: Fix typo in documentation comment. Sun Nov 25 03:58:21 GMT 2007 Olly Betts <olly@survex.com> * common/remoteserver.h,matcher/localmatch.cc,matcher/multimatch.cc: Remove lingering traces of StatsGatherer. Sat Nov 24 20:44:03 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/stats.h,matcher/Makefile.mk,matcher/stats.cc: Add get_description() method for Xapian::Weight::Internal, and add stats.cc back to hold it's implementation. * common/output.h: Add an output function for Stats (aka Xapian::Weight::Internal) to fix the build when debug logging is enabled. Sat Nov 24 01:21:56 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * api/omenquire.cc,common/,matcher/,net/remoteserver.cc: Remove StatsGatherer and its subclasses completely. Instead of a StatsGatherer subclass, MultiMatch is now simply passed a Stats object, and passes this to the prepare_match() methods of the sub matchers to be populated with statistics. OmEnquire then passes the same Stats object to MultiMatch::get_mset(), whereas a remote submatch sends the Stats object to the parent match, to be combined with the global statistics, and then passes the returned global statistics to MultiMatch::get_mset(). Remove stats.cc since it only contained implementations of StatsGatherer methods, and networkstats.h which only contained the definition of the NetworkStatsGatherer. * matcher/rset.cc: Add the rset size to the database - we now compute the global rset size the same way as we compute the global collection size: by adding the sizes in the sub collections together. We have the total rset size easily available in the code, but building the total rset size up from the sub-databases makes the code simpler. Sat Nov 24 01:21:18 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/andnotpostlist.cc: Add a missing #include "omdebug.h". Fri Nov 23 23:13:12 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/stats.h,matcher/: Remove the StatsSource class entirely. LocalSubMatch and RemoteSubMatch now contribute their stats directly to the gatherer in the prepare_match() method, so the gatherer doesn't need to worry about keeping track of the sources of statistics. Fri Nov 23 21:46:32 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/submatch.h,matcher/: Change SubMatch::start_match() (and all start_match() methods in subclasses) to take the statistics for the whole collection as an additional parameter. MultiMatch now gets the statistics from the gatherer and passes it to start_match(). Fri Nov 23 17:38:28 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/stats.h: Add set_termfreq() and set_reltermfreq() convenience methods to Stats; and remote the take_my_stats(), my_termfreq_is() and my_reltermfreq_is() methods from StatsSource. * common/rset.h,matcher/rset.cc: Rename give_stats_to_statssource() to contribute_stats(), and give it take a Stats object instead of a StatsSource object. Can't forward declare Stats, so we need to #include "stats.h" in rset.h now, unfortunately. * matcher/localmatch.cc,matcher/localmatch.h: Move register_term() implementation inline into prepare_match(). Build up the statistics in a local Stats object, rather than passing them piecemeal to the StatsSource object, and then use StatsSource::get_my_stats() to set them, just like RemoteSubMatcher. Fri Nov 23 14:35:49 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/localmatch.cc,matcher/localmatch.h: Get the total statistics from the gatherer directly and store them in a new member of LocalSubMatch, rather than getting them via the StatsSource when we need them. This bypasses the get_total_stats() and set_total_stats() methods of StatsSource. * common/stats.h,matcher/stats.cc: Remove the (newly) unused methods StatsSource::set_total_stats() and StatsSource::get_total_stats(). Fri Nov 23 11:40:54 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/,include/xapian/enquire.h,matcher/: Rename Xapian::Weight::Internal to StatsSource (which I hope to be able to remove shortly), and rename the Stats class to Xapian::Weight::Internal. Requires minimal plumbing changes - TradWeight and BM25Weight classes now query the internals directly for the statistics, LocalSubMatch now gets the statistics from StatsSource and passes them to wt_factory. common/remote-database.h and common/serialise.h now need to #include stats.h unfortunately, because a forward declaration can't be used for Xapian::Weight::Internal (unless I'm missing something). Fri Nov 23 10:15:43 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/stats.h,matcher/bm25weight.cc,matcher/localmatch.cc, matcher/tradweight.cc: Remove the Xapian::Weight::Internal get_total_* methods: instead, add a get_total_stats() method which returns a Stats object representing the whole collection. Add get_termfreq() and get_reltermfreq() convenience methods to Stats for looking up individual term's statistics. Also, change protected members of Xapian::Weight::Internal to private members, since we're no longer subclassed. Also, in tradweight.cc, avoid division by zero in the (rare) situation of the average length being zero; in the same way as this is handled by bm25weight.cc. Thu Nov 22 12:58:02 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/stats.h,matcher/localmatch.cc,matcher/localmatch.h, matcher/stats.cc: Replay Xapian::Weight::Internal's perform_request() method with set_total_stats(). The replacement takes the stats as a parameter, rather than magically getting them from the gatherer. The LocalSubMatch now needs to call the gatherer to get the stats, and then pass it to set_total_stats(), so needs to keep a reference to the gatherer. However, we are now one step closer to decoupling the gatherer from the stats. Thu Nov 22 11:47:56 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/stats.h,matcher/localmatch.cc: Remove automatic just-in-time calls to statssource.perform_request() - instead, make it public, and call it explicitly before the postlist tree is constructed. Also, make Xapian::Weight::Internal's destructor non-virtual, since it's no longer subclassed anywhere. Thu Nov 22 11:02:39 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/networkstats.h,matcher/stats.cc: Remove fetch_local_stats() method, and use of have_gathered flag for caching result of gathering the stats - instead, simply Assert that have_gathered is false before gathering; we currently only call get_local_stats() once for a given gatherer. Thu Nov 22 10:04:40 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/rset.h: Change multiline "///" comments to use "/**", style. Thu Nov 22 09:27:42 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/localmatch.cc,matcher/queryoptimiser.h: Update a couple of comments. Thu Nov 22 08:29:19 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/networkstats.h,common/stats.h,matcher/: Remove the LocalStatsSource and NetworkStatsSource classes - just use the base class (Xapian::Weight::Internal) instead. LocalStatsSource had no extra members whatsoever, whereas NetworkStatsSource had various members which weren't used, and a single method "take_remote_stats" which simply sets the statistics held - I've renamed this method "set_my_stats" and added it to the base class. Thu Nov 22 00:05:06 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/networkstats.h,common/stats.h,matcher/stats.cc: Replace StatsGatherer::contrib_my_stats() and StatsSource::contrib_stats() methods; instead add a get_my_stats() method to StatsSource and simply call this. Wed Nov 21 16:37:10 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/networkstats.h,common/remoteserver.h,matcher/stats.cc, net/remoteserver.cc: Rework NetworkStatsGatherer - it no longer needs to have the RemoteServer passed into it; instead, the RemoteServer calls NetworkStatsGatherer::set_global_stats() with the global statistics as soon as it gets them. This new method replaces fetch_global_stats(). As a result, the RemoteServer no longer needs to remember the global_stats received from the network - it just passes them straight to the gatherer - so remove the global_stats and get_global_stats() methods from RemoteServer. Wed Nov 21 15:42:40 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/rset.h,matcher/rset.cc: Document the member functions of RSetI, and remove the totally unused "get_reltermfreq()" member. Also remove an ancient, commented out, implementation of an "add_document()" member, and add copyright lines. Wed Nov 21 15:26:11 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/rset.h,matcher/localmatch.cc,matcher/rset.cc: Make RSet::calculate_stats() private, and call it only from RSet::give_stats_to_statssource(). Was previously called directly from localmatch.cc, but this way it's clear that it's only called once. Wed Nov 21 14:30:17 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/remotesubmatch.h: Move implementation of constructor into .cc file (accidentally missed this file from the last-but-one commit, which did the necessary changes to the .cc file). Wed Nov 21 14:28:48 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/rset.cc: Improve debugging messages. Wed Nov 21 14:26:44 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/remotesubmatch.cc: Add some debugging, and reorder the header includes. Wed Nov 21 13:21:35 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: Add DEBUGCALL_STATIC macros to the new functions, and tidy some code in the constructor slightly. Wed Nov 21 10:00:48 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: Split the loop which calls prepare_match() out of the constructor into a separate static function. Tue Nov 20 18:23:53 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: Refactor code which splits the RSet into sub RSets (one for each sub database) into a separate function. No functional change. Tue Nov 20 17:56:47 GMT 2007 Olly Betts <olly@survex.com> * common/remoteserver.h: Remove unused '#include "omassert.h"'. Tue Nov 20 12:53:06 GMT 2007 Olly Betts <olly@survex.com> * AUTHORS: Thank Marcus Rueckert for GCC 4.3 fixes. Tue Nov 20 10:22:28 GMT 2007 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Recent GCC 4.3 snapshots thinks cf and tf might be used uninitialised. They won't be, but we only need to initialise them once per run to silence the warning, so just do that. Tue Nov 20 10:20:13 GMT 2007 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Fix equality testing of C strings to use strcmp() rather than '=='. In practice, using '==' often gives the desired effect due to pooling of constant strings, but this may have resulted in bugs on some platforms. Tue Nov 20 01:47:16 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/backendmanager_remotetcp.cc: Add missing "#include <cstring>". Tue Nov 20 01:26:40 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/unixcmds.cc: Add #include of cstring, to get strchr, needed for gcc 4.3 snapshot. Tue Nov 20 01:25:22 GMT 2007 Olly Betts <olly@survex.com> * api/maptermlist.h,api/termlist.cc,backends/alltermslist.cc, backends/flint/flint_spelling.cc,backends/flint/flint_spelling.h, common/alltermslist.h,common/termlist.h,common/vectortermlist.h: Provide a default implementation of accumulate_stats() in the virtual base class TermIterator::Internal instead of repeating it in each subclass which doesn't get used for generating an ESet, and don't call abort() in the default implementation - an Assert(false) is sufficient, and more consistent with how we handle other similar cases. Tue Nov 20 01:24:17 GMT 2007 Olly Betts <olly@survex.com> * bin/xapian-progsrv.cc: Add missing '#include <cstdlib>'. Mon Nov 19 23:54:53 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_postlist.cc,backends/quartz/quartz_postlist.cc: Add NORETURN macro to report_read_error(); fixes warnings from GCC 4.3 about possibly uninitialised values. Reorder header includes to follow proposed policy. * backends/flint/flint_postlist.h: Add include of omdebug.h which previously needed to be done before including this. Tidy up order of includes. Mon Nov 12 14:06:38 GMT 2007 Olly Betts <olly@survex.com> * api/Makefile.mk,api/omenquire.cc,api/sorter.cc,common/multimatch.h, common/omenquireinternal.h,include/Makefile.mk,include/xapian.h, include/xapian/enquire.h,include/xapian/sorter.h, matcher/multimatch.cc,net/remoteserver.cc,tests/: Add new functor class to allow more sophisticated sorting options. Mon Nov 12 14:03:06 GMT 2007 Olly Betts <olly@survex.com> * HACKING: Improved wording. Sun Nov 11 07:51:57 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Hide the v102 namespace from Doxygen as it isn't user visible. Sat Nov 10 21:53:21 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am: Remove .multi in clean-local rule; should fix distcheck. Sat Nov 10 15:34:25 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/testutils.cc,tests/harness/testutils.h: mset_expect_order_begins() is never used, so remove it. * tests/harness/testutils.cc: Don't need <iostream>, only <fstream>. Sat Nov 10 11:09:22 GMT 2007 Olly Betts <olly@survex.com> * docs/index.html: Add a list of documents on particular features and include links to previously unlinked-to documents. Weed down the top navigation bar which had grown to unwieldy length. Sat Nov 10 10:17:42 GMT 2007 Olly Betts <olly@survex.com> * PLATFORMS: Update for Debian buildds. Sat Nov 10 02:13:07 GMT 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: If "first" is non-zero, then use nth_element with vector::reverse_iterator to partition items such that the unwanted ones are at the end not the start. This means the call to vector::erase() no longer needs to copy all the wanted items. Fri Nov 09 19:01:38 GMT 2007 Olly Betts <olly@survex.com> * tests/api_db.cc: collapsekey3 is too strict - even if the value never occurs, lower_bound may drop as we may reject potential matches before applying the collapse test. Fri Nov 09 16:36:39 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/termiterator.h: Clarify get_wdf() versus get_termfreq() in documentation comments. Fri Nov 09 00:07:02 GMT 2007 Olly Betts <olly@survex.com> * tests/api_db.cc: Fix comment typo. Fri Nov 09 00:04:45 GMT 2007 Olly Betts <olly@survex.com> * HACKING,Makefile.am,tests/Makefile.am: New make target 'check-multi'. Thu Nov 08 23:54:47 GMT 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Re-enable test case allpostlist1 for multi since it works fine (presumably thanks to the multi_postlist.cc fix). Thu Nov 08 23:44:29 GMT 2007 Olly Betts <olly@survex.com> * tests/,tests/harness/: Add new "multi" test backend which indexes the specified text file(s) to two databases which will look just like the equivalent single database when searched together. Since writing isn't supported, a number of tests which require this are disabled, as are any tests which explicitly use a multi-database since a multi-of-multis isn't equivalent to a multi of equivalent individual databases. A handful of testcases are also currently disabled because they fail for reasons not yet fully explored. Existing tests run with the new backend provide regression tests for the two recent fixes. Thu Nov 08 23:42:33 GMT 2007 Olly Betts <olly@survex.com> * backends/multi/multi_postlist.cc: Fix PostingIterator::skip_to() when running over multiple databases. Regression test to follow. Thu Nov 08 23:37:51 GMT 2007 Olly Betts <olly@survex.com> * docs/overview.html: Mention the "auto" backend. Thu Nov 08 23:28:30 GMT 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix bug in handling a pure boolen match over more than one database under set_docid_order(ASCENDING) - we can't shortcut in this case because MergePostList generally doesn't return docids in order. Regression test to follow. Thu Nov 08 21:07:09 GMT 2007 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Fix comment typo. Thu Nov 08 07:40:48 GMT 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc: Don't use MultiPostList when there's only one subdatabase. Thu Nov 08 05:40:17 GMT 2007 Olly Betts <olly@survex.com> * tests/api_db.cc: Run sortrel1 for inmemory too. Thu Nov 08 03:43:50 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc,tests/harness/index_utils.cc, tests/harness/index_utils.h: New class FileIndexer which acts like an iterator returning a Xapian::Document object for each paragraph in some specified files. Wed Nov 07 16:30:49 GMT 2007 Olly Betts <olly@survex.com> * common/,docs/remote_protocol.html,matcher/,net/remoteserver.cc, net/serialise.cc: If we're doing a match with only one database which is remote then just return the unserialised MSet from the remote match. This requires that we include internal->percent_factor in the MSet serialisation, which requires a minor remote protocol version bump. Wed Nov 07 09:47:28 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: Fix logging build (a change to the logging code introduced in r9644 didn't compile) Tue Nov 06 17:10:51 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/document.h: Better documentation comment for Document::termlist_count(). Tue Nov 06 12:32:16 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,docs/Makefile.am,languages/Makefile.mk: No need to set SUFFIXES manually for suffixes used in implicit rules. Tue Nov 06 10:41:14 GMT 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: If we're collapsing on a value, keep track of the number of empty collapse values seen, since that allows us to give a better lower bound on the number of matches. * tests/api_db.cc: This change breaks an (incorrect) assumption in collapsekey3, so change that test case to be a regression test for the improved estimate. Add a new test as (the previously missing) collapsekey2, but disable it for now as we don't seem to have a suitable existing database. Tue Nov 06 07:58:34 GMT 2007 Olly Betts <olly@survex.com> * matcher/tradweight.cc: Make sure lenpart has been calculated in TradWeight::get_sumpart_needs_doclength() (this doesn't appear to actually be a problem in current use, but it might become an issue if the code which uses TradWeight changes. Tue Nov 06 07:49:18 GMT 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Add simple feature test for TradWeight being used to run a query. Tue Nov 06 07:26:28 GMT 2007 Olly Betts <olly@survex.com> * common/stats.h,matcher/bm25weight.cc,tests/api_anydb.cc: Fix bug in BM25Weight - in the case where k2 is non-zero, a non-initialised value influenced the weight calculations. By default k2 is zero, so this bug probably won't affect most users. Tue Nov 06 07:22:27 GMT 2007 Olly Betts <olly@survex.com> * api/omenquire.cc,common/multimatch.h,matcher/multimatch.cc, net/remoteserver.cc: Modify MultiMatch::MultiMatch() to take a const pointer to Xapian::RSet rather than a const reference. This makes the code simpler for the case where we don't have an RSet. Tue Nov 06 05:47:38 GMT 2007 Olly Betts <olly@survex.com> * api/omenquire.cc: Use ".empty()" not ".size() == 0". Mon Nov 05 21:05:45 GMT 2007 Olly Betts <olly@survex.com> * m4/xapian.m4: Hook LT_INIT as well as AC_PROG_LIBTOOL and AM_PROG_LIBTOOL. Mon Nov 05 11:17:23 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * docs/termgenerator.rst: Change a couple of instances of "terms" to "words" for clarity; the item from the input text is a word, and the result of processing is a term. Mon Nov 05 09:46:53 GMT 2007 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Note that this document is up-to-date for 1.0.4. Improve the formatting a bit and fix a the odd typo. Mon Nov 05 06:24:01 GMT 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Fix MSVC warning. Mon Nov 05 06:21:09 GMT 2007 Olly Betts <olly@survex.com> * matcher/queryoptimiser.h: Forward declare PosFilter as "struct" rather than "class", since it's defined as "struct". Mon Nov 05 05:41:01 GMT 2007 Olly Betts <olly@survex.com> * api/sortable-serialise.cc: Suppress MSVC warning. Mon Nov 05 04:55:48 GMT 2007 Olly Betts <olly@survex.com> * docs/quickstart.html: Improved. Mon Nov 05 04:43:13 GMT 2007 Olly Betts <olly@survex.com> * docs/overview.html: More improvements from Jenny Black. Sun Nov 04 22:59:28 GMT 2007 Olly Betts <olly@survex.com> * docs/overview.html: Assorted improvements to the start of this document. Sun Nov 04 22:54:13 GMT 2007 Olly Betts <olly@survex.com> * api/omdocument.cc,tests/api_wrdb.cc: values_begin() didn't ensure that values had been read. However, values_end() did (and so did values_count()) so this wasn't generally an issue, but it shouldn't happen anyway. * tests/api_wrdb.cc: Extend adddoc5 to include a regression test for this bug. Sun Nov 04 22:44:39 GMT 2007 Olly Betts <olly@survex.com> * tests/quartztest.cc: Remove 'FIXME' suggesting more value tests. * tests/api_nodb.cc: Add the suggested tests we don't already have. Sun Nov 04 19:23:52 GMT 2007 Olly Betts <olly@survex.com> * tests/api_posdb.cc,tests/api_wrdb.cc: Use get_writable_database() instead of get_writable_database(""). * tests/api_wrdb.cc: Rework test_spell3 so it works under __WIN32__ (bug#177). Sun Nov 04 07:31:04 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.h: Make BackendManager dtor virtual as intended. Sun Nov 04 07:05:12 GMT 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc,tests/apitest.cc,tests/apitest.h,tests/harness/, tests/quartztest.cc: Extend BackendManager to allow opening a WritableDatabase as a Database as well, and to allow reopening a closed WritableDatabase as a WritableDatabase. With these new features, rework quartztest's adddoc2 and adddoc3 as apitest testcases adddoc5 and adddoc6. Sun Nov 04 06:08:42 GMT 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc,tests/apitest.cc,tests/apitest.h,tests/harness/: Add BackendManager::get_writable_database_as_database() and use it instead of assuming what the WritableDatabase path will be. Sun Nov 04 05:31:30 GMT 2007 Olly Betts <olly@survex.com> * tests/,tests/harness/: Subclass BackendManager for each database backend supported, which should provide a better foundation on which we can rebuild to fix the nastier bits of the test harness. Fri Nov 02 17:39:58 GMT 2007 Olly Betts <olly@survex.com> * INSTALL: zlib 1.2.0 apparently fixes a memory leak in deflateInit2, which we use, so that's another reason to prefer 1.2.x. Fri Nov 02 06:34:36 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Fix extracting of valgrind error messages. Fri Nov 02 05:09:05 GMT 2007 Olly Betts <olly@survex.com> * docs/bm25.html,docs/scalability.html: Remove references to Muscat 3.6. Thu Nov 01 18:51:20 GMT 2007 Olly Betts <olly@survex.com> * NEWS: Fix to talk about OP_VALUE_RANGE rather than ValueRangePostList since the later isn't user-visible. Thu Nov 01 18:16:20 GMT 2007 Olly Betts <olly@survex.com> * matcher/exactphrasepostlist.cc: Fix memory leak if second memory allocation fails. Thu Nov 01 17:54:09 GMT 2007 Olly Betts <olly@survex.com> * docs/intro_ir.html: Move the section on stemming to a better location. Thu Nov 01 17:37:40 GMT 2007 Jenny Black * docs/glossary.rst: Fix typo. Thu Nov 01 06:34:58 GMT 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Minor code simplification. Wed Oct 31 16:17:08 GMT 2007 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Fix to handle the special case of a term which is 48 characters long. Fix not to go into an infinite loop if certain checks fail. Wed Oct 31 15:56:45 GMT 2007 Olly Betts <olly@survex.com> * matcher/multiandpostlist.cc: If the subpostlists are ORs and pruning or operator decay happens within them, then get_termfreq_est() can change such that the first postlist returns a higher value than the second, so remove a bogus assertion which assumed this didn't happen (bug#209). Wed Oct 31 15:43:58 GMT 2007 Olly Betts <olly@survex.com> * backends/flint/flint_version.cc: I documented an increase in FLINT_VERSION but didn't increase it - luckily the previous version wasn't in a released version, so just adjust the comment to match what actually happened. Wed Oct 31 15:24:48 GMT 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Add '#include <safeunistd.h>' (bug#208). Wed Oct 31 15:01:53 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-check.cc: Update the checking of the termlist table to stop checking for the "has_termfreqs" flag, which is no longer stored in termlists. Wed Oct 31 14:31:38 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * backends/quartz/quartz_positionlist.cc: Add missing #include, which is needed only when configured with --enable-log and with --disable-assertions. Tue Oct 30 04:52:53 GMT 2007 Olly Betts <olly@survex.com> * NEWS,configure.ac: Update for 1.0.4. Tue Oct 30 04:10:32 GMT 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Code tweak. Tue Oct 30 03:04:04 GMT 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add two more test cases to improve coverage. Tue Oct 30 00:13:42 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: abi::__cxa_demangle() requires GCC 3.1 it seems. Mon Oct 29 23:14:41 GMT 2007 Olly Betts <olly@survex.com> * docs/scalability.html: Update size of gmane. Mon Oct 29 23:13:00 GMT 2007 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Note that Quartz is now deprecated. Mon Oct 29 21:09:12 GMT 2007 Olly Betts <olly@survex.com> * configure.ac,tests/harness/testsuite.cc: Demangle the name of the std::exception subclass for GCC, and handle compilation with -fno-rtti. Mon Oct 29 19:50:05 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Catch std::exception subclasses explicitly and report e.what() (which gives the method name which threw, at least under GCC). Mon Oct 29 18:46:57 GMT 2007 Olly Betts <olly@survex.com> * NEWS: Updated. Mon Oct 29 18:36:12 GMT 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Pull out old_item.wt into a variable. Mon Oct 29 18:24:46 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.lemony: Fix problem with spelling correction of hyphenated terms (or other terms joined with phrase generators): the position of the start of the term wasn't being reset for the second term in the generated phrase, resulting in out of bounds errors when substituting the new value in the corrected query string. * tests/queryparsertest.cc: Test the fix. Mon Oct 29 17:34:17 GMT 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Move "Stem::stem_word(word)" in the bindings to the right section (it was done in 1.0.0, as already indicated). Deprecate the non-pythonic iterators in favour of the pythonic ones. Mon Oct 29 02:54:09 GMT 2007 Olly Betts <olly@survex.com> * NEWS,api/Makefile.mk,docs/Makefile.am,include/Makefile.mk, include/xapian.h,tests/api_db.cc,tests/api_nodb.cc, tests/api_wrdb.cc: Back out match spy changes in preparation for creating a branch for them. Mon Oct 29 02:37:31 GMT 2007 Olly Betts <olly@survex.com> * matcher/valuerangepostlist.cc: Fix ValueRangeProcessor::get_termfreq_est() and get_termfreq_max() to work when at_end() when paranoid assertions are enabled. Mon Oct 29 01:51:07 GMT 2007 Olly Betts <olly@survex.com> * common/omenquireinternal.h: Add ESetItem::swap() and MSetItem::swap(). Sun Oct 28 16:33:34 GMT 2007 Olly Betts <olly@survex.com> * PLATFORMS: Update from tinderbox and buildbot. Sun Oct 28 06:00:23 GMT 2007 Olly Betts <olly@survex.com> * matcher/queryoptimiser.cc: Remove unused variable. Sun Oct 28 05:46:28 GMT 2007 Olly Betts <olly@survex.com> * docs/remote.html: xapian-tcpsrv can handle concurrent read access so update the out-of-date information here. Also, some new features aren't supported by the remote backend yet. Sun Oct 28 05:20:02 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Update documentation comment for QueryParser::set_stemming_strategy(). Sun Oct 28 05:03:03 GMT 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix handling of STEM_ALL. * tests/queryparsertest.cc: Update tests. Move tables of queries to before the test functions which use them. Sat Oct 27 20:50:57 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,include/xapian/query.h, matcher/queryoptimiser.cc: Eliminate Query::Internal::dbl_parameter to avoid any risk of ABI breakage. Sat Oct 27 17:08:46 BST 2007 Olly Betts <olly@survex.com> * include/xapian/query.h: Pull in <xapian/deprecated.h> explicitly. Add @deprecated note to Query::Query(Query::op, Query) explaining why it is deprecated. * include/xapian/queryparser.h: Wrap comment better. Sat Oct 27 05:37:33 BST 2007 Olly Betts <olly@survex.com> * NEWS: Merge two "testsuite" sections in draft 1.0.4 entry. Sat Oct 27 02:34:10 BST 2007 Olly Betts <olly@survex.com> * HACKING: Note specifically that std::list::size() is O(n) for GCC. Update the debian packaging checklist. Sat Oct 27 02:33:14 BST 2007 Olly Betts <olly@survex.com> * docs/intro_ir.html: Add link to the forthcoming book "Introduction to Information Retrieval", which can be read online. Sat Oct 27 02:22:36 BST 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: We no longer build sourcedoc.pdf so remove it from MAINTAINERCLEANFILES. Sat Oct 27 02:20:31 BST 2007 Olly Betts <olly@survex.com> * NEWS: Mostly updated for 1.0.4. Fri Oct 26 04:57:31 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: OP_SCALE_WEIGHT applied to MatchNothing is now handled by the Query ctor, so replace the code to handle OP_SCALE_WEIGHT in Query::Internal::simplify_matchnothing() by an assertion that the subquery isn't MatchNothing. Fri Oct 26 04:50:37 BST 2007 Olly Betts <olly@survex.com> * include/xapian/query.h: Deprecate the essentially useless constructor Query(Query::op, Query). * docs/deprecation.rst: Note this deprecation, and since the feature doesn't actually have a use, schedule it for 1.1.0. Also schedule Enquire::register_match_decider() for 1.1.0 for the same reason. Fri Oct 26 01:00:08 BST 2007 Olly Betts <olly@survex.com> * api/omquery.cc,include/xapian/query.h: OP_SCALE_WEIGHT applied to OP_VALUE_RANGE can have no effect so ignore it. * tests/queryparsertest.cc: Update expected query descriptions. Fri Oct 26 00:23:55 BST 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: When checkatleast is set, and we're sorting by relevance with forward ordering by docid, and the query is pure boolean, we were exiting before the checkatleast requirement was satisfied. Then the adjustments made to the estimated and max statistics based on checkatleast meant we claimed there were exactly msize results. Thu Oct 25 00:34:14 BST 2007 Olly Betts <olly@survex.com> * matcher/localmatch.cc,matcher/multiandpostlist.cc, matcher/queryoptimiser.cc,queryparser/queryparser.cc: Fix build with --enable-assertions and --enable-log. Wed Oct 24 06:29:03 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Drop out of IN_GROUP mode when we generate a BOOLEAN_FILTER token. * tests/queryparsertest.cc: Add regression test. Wed Oct 24 06:08:29 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Drop special treatment for unmatched ')' at the start of the query, as it seems rather arbitrary and not particularly useful. Reparsing seems a better approach here. * tests/queryparsertest.cc: Add regression test case for bug which dropping this special case fixes. Wed Oct 24 05:25:40 BST 2007 Olly Betts <olly@survex.com> * AUTHORS: Add Ron Kass for several bug reports. Wed Oct 24 05:13:01 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix parsing of queries which consist only of boolean filter terms and HATE-d terms. * tests/queryparsertest.cc: Add regression tests, and a few more cases which passed before too. Fix descriptions of queries which now give different (but equivalent) Xapian::Query object hierarchies. Wed Oct 24 04:51:56 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Fix warning in debug log build. Wed Oct 24 04:41:08 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc,backends/flint/flint_table.h: Don't use class member function pointers to implement FlintTable::next() and FlintTable::prev() as we can just call the methods directly based on the sequential flag. Profiling suggests this speeds up searches a little, the two pointers take up 16 bytes each (on x86_64) so it reduces the class size by up to 32 bytes, and the code is simpler because we no longer need to update prev_ptr and next_ptr. Wed Oct 24 00:16:06 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Tweak Query::get_description() to describe OP_SCALE_WEIGHT queries as '<factor> * <subquery>' rather than '<subquery> * <factor>' as the former results in more readable descriptions. * queryparser/queryparser.lemony: Use OP_SCALE_WEIGHT with factor 0.0 for queries which should be purely boolean which resolves a FIXME comment in the code. * tests/queryparsertest.cc: Fix expected results for this change. Tue Oct 23 19:13:08 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Some as_XXX() methods delete the object they are called on, while others don't, so rename the ones which don't to get_XXX() to make the code clearer, and make them all return Query rather than Query *. Eliminate Term::as_query() completely - when we actually want a new object we can just call new in the caller. op_window_query() now takes the difference between the number of terms and the total window size, and we now use a std::vector to store the generated Query objects rather than a std::list, since we know how many there will be and can reserve the required size in advance. Tue Oct 23 18:33:49 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Improve comment. Mon Oct 22 21:00:05 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix handling of LOVE and HATE following a quoted phrase. * tests/queryparsertest.cc: Add regression test. Mon Oct 22 05:23:23 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Refactor so Term::as_query() calls Term::as_query_object() instead of vice versa, as this avoids calling 'new Query' quite a bit and makes a measurable difference to the speed of the QueryParser. Fix a call to as_query() which can be as_query_object(). Mon Oct 22 04:10:17 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: FLAG_PARTIAL with multi-prefixes would result in inflated wqf for the "normal" version of the term treated as partial. * tests/queryparsertest.cc: Add regression test. Mon Oct 22 02:34:58 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,queryparser/queryparser_internal.h: Rename QueryParser::Internal::prefixes to prefixmap to avoid confusion with all the other variables which are now called "prefixes" in queryparser.lemony. Eliminate have_prefix and instead just set prefixinfo to NULL by default. Sun Oct 21 23:56:57 BST 2007 Olly Betts <olly@survex.com> * common/omenquireinternal.h,include/xapian/enquire.h, include/xapian/query.h,matcher/: Implement a new QueryOptimiser class which can hoist the positional filters of OP_PHRASE and OP_NEAR higher up the tree and merge the "AND" inside them into any neighbouring OP_AND or OP_FILTER (bug#23). This shaves 10% of the execution time of real world queries on real world data. On particularly slow cases, the saving can be more dramatic - it saves 50% when tested on a log of slow cases. OP_SCALE_WEIGHT scaling factors are now pushed down to the leaves and any leaf we a factor other than 0.0 (boolean) or 1.0 (unscaled) has its weights scaled using a ScaleWeight wrapper around the normal weighting object (bug#203). Sun Oct 21 23:46:11 BST 2007 Olly Betts <olly@survex.com> * include/xapian/query.h: Fix documentation of OP_SCALE_WEIGHT. Negative scaling factors aren't now clipped to 0, instead we throw Xapian::InvalidArgumentError. Remove the explicit setting of OP_ELITE_SET to 10 which is no longer required. Wrap a long comment line. Sun Oct 21 23:20:23 BST 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Tweak code and wrap comment. Sun Oct 21 04:53:40 BST 2007 Olly Betts <olly@survex.com> * matcher/extraweightpostlist.h: Add missing '#include "multimatch.h"' which is pulled in implicitly by some other header which always happens to be included before this one. Fri Oct 19 03:52:17 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Remove deprecation of QueryParser::add_prefix() and QueryParser::add_boolean_prefix(). Reformat the tables to use the "simple table" style, and to all be 126 columns wide, since that fits in a maximised terminal window without wrapping (at least on my machine!) Fri Oct 19 03:44:33 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.cc, queryparser/queryparser.lemony,queryparser/queryparser_internal.h, tests/queryparsertest.cc: Since calling QueryParser::add_prefix() or QueryParser::add_boolean_prefix() a second time with the same field name was ignored before (rather than overriding as we had thought) it seems reasonable to change this behaviour. This also avoids the need to deprecate these methods which will force all users to update their code. Change the semantics of default_prefix - it's more useful if this overrides any default prefixes set with add_prefix(), since this allows the same QueryParser object to parse both a "general" search input which searches several fields and also a "title" search input. Trying to set the same field as probabilistic and boolean now throws InvalidOperationError rather than UnimplementedError. Fri Oct 19 03:29:53 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Fix compilation. Fri Oct 19 03:28:37 BST 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: We're still getting "info" messages in the generated HTML. It seems to be impossible to get rst2html to report "info" messages to stdout without adding them to the generated document, so just drop "--verbose" to disable them completely. Thu Oct 18 18:27:28 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Use std::vector<> for the stack in the lemon-generated parser. This means that the stack is no longer a fixed size, so we needn't worry about overflow, and that in typical use it'll actually use less memory (lemon defaults to a 100 entry stack). Thu Oct 18 17:23:45 BST 2007 Olly Betts <olly@survex.com> * languages/turkish.sbl: Minimise differences with Snowball SVN HEAD by stripping trailing whitespace. Mon Oct 15 15:39:26 BST 2007 Olly Betts <olly@survex.com> * backends/quartz/btree_util.h: Need "safeunistd.h" for close(). Fixes build errors on mingw and with SGI's CC on IRIX. Mon Oct 15 05:30:23 BST 2007 Olly Betts <olly@survex.com> * matcher/localmatch.cc: Fix typo in comment. Mon Oct 15 05:26:35 BST 2007 Olly Betts <olly@survex.com> * common/omassert.h: Rewritten from scratch. The new version only includes headers if assertions are enabled, which should help to speed up non-assertion builds by reducing unnecessary header inclusion. Also, float.h and math.h are never now pulled in - instead we use the new within_DBL_EPSILON() function. AssertNe() and AssertNeParanoid() are never actually used, so replace them with AssertRel() and AssertRelParanoid which allow the user to assert any binary relation, not just inequality. Also, we now use rare() to give branch prediction hints for assertion tests (since the failure branch should never be taken). * common/omdebug.h,common/stringutils.h,tests/harness/testsuite.h: Replace several definitions of the STRINGIZE macro with a single version in common/stringutils.h. * backends/flint/,backends/inmemory/inmemory_database.cc, backends/multi/multi_postlist.cc,backends/quartz/, backends/remote/remote-database.cc,bin/quartzcheck.cc, bin/xapian-compact.cc,common/stringutils.h,expand/expandweight.cc, expand/ortermlist.cc,matcher/phrasepostlist.cc, matcher/scaleweightpostlist.cc,net/remoteconnection.cc, net/tcpserver.cc: Explicitly include headers which were previously being pulled in implicitly by omassert.h. * HACKING: Update the documentation for assertion calls, and document CompileTimeAssert() (which previously wasn't documented here). Mon Oct 15 05:09:15 BST 2007 Olly Betts <olly@survex.com> * common/utils.cc,common/utils.h: Add within_DBL_EPSILON() function which returns true if its two double arguments differ by less than DBL_EPSILON (currently not used anywhere). Mon Oct 15 05:03:50 BST 2007 Olly Betts <olly@survex.com> * matcher/multiandpostlist.cc: If check sets valid to true, we can't be at_end(), so check valid first as at_end() is a rare event. Mon Oct 15 05:00:18 BST 2007 Olly Betts <olly@survex.com> * matcher/: Remove code for FilterPostList, which has been unused for ages. Sun Oct 14 02:13:57 BST 2007 Olly Betts <olly@survex.com> * matcher/andpostlist.cc: AndPostList now ensures that its left is less frequent than its right (it can still be produced as an operator decay product, and sometimes left is more frequent when this happens). Sun Oct 14 00:58:20 BST 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Use rare() to mark rarely taken branches. Sun Oct 14 00:30:46 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Add rare() and usual() macros to config.h which allow branch prediction hints to be given for compilers which support this (currently GCC and Intel C++). * HACKING: Document rare() and usual(). Sat Oct 13 22:37:32 BST 2007 Olly Betts <olly@survex.com> * HACKING: Improve wording. Sat Oct 13 16:42:26 BST 2007 Olly Betts <olly@survex.com> * api/postlist.cc,common/postlist.h,matcher/: Implement a variant of PostList::skip_to() called PostList::check() which isn't required to leave the PostList on a particular docid. This allows queries filtered by a ValueRangePostList to run around 3.5 times faster. Fixes buf#164, though there's probably scope for at least some further improvement. Sat Oct 13 16:08:07 BST 2007 Olly Betts <olly@survex.com> * matcher/multiandpostlist.cc: Fix typo bug (get_termfreq_min should be get_termfreq_max!) Sat Oct 13 15:53:52 BST 2007 Olly Betts <olly@survex.com> * matcher/multiandpostlist.cc: Fix assertion (should be <= not <). Fri Oct 12 01:20:05 BST 2007 Olly Betts <olly@survex.com> * HACKING: Note that rst2html may be installed as rst2html.py. Fri Oct 12 01:10:11 BST 2007 Olly Betts <olly@survex.com> * matcher/localmatch.cc: Tweak formatting. Thu Oct 11 22:58:36 BST 2007 Olly Betts <olly@survex.com> * tests/: svn:ignore: Add termgentest and termgentest.exe. Thu Oct 11 16:18:03 BST 2007 Olly Betts <olly@survex.com> * configure.ac: If rst2html isn't found, also look for rst2html.py, which archlinux reportedly installs it as. Thu Oct 11 16:09:20 BST 2007 Olly Betts <olly@survex.com> * matcher/: Round the result of all get_termfreq_est() calculations to the nearest integer instead of rounding down. Thu Oct 11 15:40:48 BST 2007 Olly Betts <olly@survex.com> * matcher/: Add new PostList subclass MultiAndPostList which handles a multi-way AND operation in a single class. This allows us to optimise some cases of 3 or more way AND operations much better and gives a 16-17% performance improvement in tests using real-world query logs. * tests/api_anydb.cc: MultiAndPostList rounds get_termfreq_est() calculations to the nearest integer (rather than always rounding down) so adjust the expected answers in test_matches. Thu Oct 11 15:14:27 BST 2007 Olly Betts <olly@survex.com> * matcher/selectpostlist.cc: Explicitly qualify next() as SelectPostList::next(). Thu Oct 11 13:04:40 BST 2007 Olly Betts <olly@survex.com> * README: Remove the ancient history lesson - this material is better left to the history page on the website. Thu Oct 11 00:33:34 BST 2007 Olly Betts <olly@survex.com> * matcher/branchpostlist.h: Fix comment typo. Thu Oct 11 00:28:29 BST 2007 Olly Betts <olly@survex.com> * api/postlist.cc,backends/flint/flint_alldocspostlist.cc, backends/flint/flint_alldocspostlist.h,common/postlist.h,matcher/: Eliminate several implementations of open_position_list and read_position_list in favour of default ones in the PostList base class which throw InvalidOperationError. Change the default get_wdf implementation to also throw InvalidOperationError. Wed Oct 10 22:27:44 BST 2007 Olly Betts <olly@survex.com> * docs/install.html: Improve the remainder. This documents now just gives a brief overview of building, suitable for most common cases, and defers to the INSTALL document in each tarball for more details. Wed Oct 10 21:51:39 BST 2007 Olly Betts <olly@survex.com> * common/leafpostlist.h: Rewrite header. * api/Makefile.mk,api/leafpostlist.cc: Add new source file for virtual, and non-trivial, non-virtual, methods of LeafPostList. * api/omdatabase.cc,backends/inmemory/inmemory_database.h, backends/multi/multi_postlist.cc,backends/multi/multi_postlist.h, backends/remote/net_postlist.h,common/emptypostlist.h: Add missing '#include "omassert.h"' which was previously pulled in implicitly via '#include "leafpostlist.h"'. * backends/inmemory/inmemory_database.h: Remove unused '#include <stdlib.h>'. * backends/multi/multi_postlist.cc,backends/multi/multi_postlist.h: MultiPostList::set_termweight() is never used, so eliminate it. * backends/remote/net_postlist.cc: Rename parameter "weight" to "min_weight" to avoid clash with new member variable of LeafPostList. * matcher/emptysubmatch.cc,matcher/localmatch.cc: Don't call LeafPostList::set_termweight() with a BoolWeight object - the default behaviour is now equivalent. Wed Oct 10 18:18:09 BST 2007 Olly Betts <olly@survex.com> * docs/install.html: Improve the first half. Wed Oct 10 16:32:57 BST 2007 Olly Betts <olly@survex.com> * common/postlist.h: Rewrite header. * api/Makefile.mk,api/postlist.cc: Add new source file for virtual methods of Xapian::PostingIterator::Internal. * api/omdatabase.cc,backends/flint/flint_postlist.h, backends/quartz/quartz_alldocspostlist.cc, backends/quartz/quartz_postlist.h,matcher/valuerangepostlist.cc: Add missing '#include "autoptr.h"' which was previously pulled in implicitly via '#include "postlist.h"'. Wed Oct 10 16:25:08 BST 2007 Olly Betts <olly@survex.com> * tests/quartztest.cc: Remove quartztest's test_postlist1 and test_postlist2. * tests/api_db.cc: Enhance test_termstats to cover part of the removed tests. Enhance test_postlist1 to check more long terms. Eliminate the helper function from test_postlist3. * tests/api_wrdb.cc: Add test_postlist7 to cover the rest of what the removed tests checked which isn't already checked elsewhere. Wed Oct 10 02:29:44 BST 2007 Olly Betts <olly@survex.com> * HACKING: Update the release checklist. Wed Oct 10 02:01:03 BST 2007 Olly Betts <olly@survex.com> * matcher/scaleweightpostlist.cc,matcher/scaleweightpostlist.h: Move virtual dtor definition out of header. Wed Oct 10 01:41:05 BST 2007 Olly Betts <olly@survex.com> * matcher/msetpostlist.cc: Avoid virtual method overhead in MSetPostList::recalc_maxweight(). Wed Oct 10 00:16:00 BST 2007 Olly Betts <olly@survex.com> * common/expandweight.h,expand/expand.cc,expand/expandweight.cc: We never actually use ExpandWeight::get_maxweight(), so remove it. Tue Oct 09 16:27:54 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_version.cc, backends/flint/flint_version.h: Don't try to upgrade a flint database if we're only reading it, or it's already the latest version. Mon Oct 08 10:14:57 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_nodb.cc: Rename test_scaleweightmatchnothing1 to test_scaleweight3 so all the scaleweight tests can be run together more easily. Add test_scaleweight4 to test that scaling by a weight close to 1 is optimised away. Sun Oct 07 12:20:14 BST 2007 Olly Betts <olly@survex.com> * include/xapian/database.h: Add explicit note that Database::get_metadata() returns an empty string when the backend doesn't support user-specified metadata, and that WritableDatabase::set_metadata() throws UnimplementedError in this case. Add note about current behaviour with multidatabases. Sun Oct 07 10:58:24 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: No need to use AutoPtr<> here now. Sun Oct 07 10:54:05 BST 2007 Richard Boulton <richard@lemurconsulting.com> * common/remoteprotocol.h: Bump the network protocol minor revision, since we've added a query operator (OP_SCALE_WEIGHT) and the server will need to be updated to understand serialised queries containing it. Sun Oct 07 01:54:43 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/omqueryinternal.cc,include/xapian/query.h,matcher/, tests/api_anydb.cc,tests/api_nodb.cc,tests/internaltest.cc: Rename OP_MULT_WEIGHT to OP_SCALE_WEIGHT and MultWeight to ScaleWeight, and move it to before OP_MULT_WEIGHT (to remove the gap in the enum). Don't do an epsilon test to compare the weight to zero (when checking if it's effectively a boolean query) - do an exact comparison instead. Raise an exception if the parameter for OP_SCALE_WEIGHT is < 0 (instead of clipping it to 0). Adjust weights accordingly. Add test_scaleweight2() to test a query with some weights multiplied by 0, and some by a different factor. Wed Oct 03 17:23:02 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/omqueryinternal.cc,tests/api_nodb.cc: Fix (and test) for a double delete if OP_MULT_WEIGHT was applied to an empty query. Mon Oct 01 13:20:39 BST 2007 Richard Boulton <richard@lemurconsulting.com> * NEWS: Update * net/remoteconnection.cc: Back out previous change - I was confused, the code was previously fine, and the change broke it. Mon Oct 01 13:13:48 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/omqueryinternal.cc: Fix access of uninitialised member when unserialising a mult-weight query. * tests/internaltest.cc: Add a regression test for unserialisationg of mult-weight queries. Mon Oct 01 13:11:24 BST 2007 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: After calling read(), check for received == 0 after checking for errors, so that if an EINTR occurs in read, we don't report EOF instead of retrying. Sun Sep 30 23:20:34 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/omquery.cc,api/omqueryinternal.cc,include/xapian/query.h: Add OP_MULT_WEIGHT operator, and a new Query constructor which takes a single subquery and a "double" parameter - currently only useful for OP_MULT_WEIGHT. Add dbl_parameter to Query::Internal, to hold this parameter. Add serialisation of OP_MULT_WEIGHT using "." to represent the operator. Drop OP_MULT_WEIGHT operators which have a parameter of 1 (or very close) when simplifying, since these have no effect. * matcher/multweightpostlist.cc,matcher/multweightpostlist.h: New files, implementing a postlist which multiplies the weights from its single sub-postlist by a parameter. * matcher/Makefile.mk: Add new files. * matcher/localmatch.cc: Add support for the OP_MULT_WEIGHT query operator, which produces a MultWeightPostList if the associated parameter is greater than DBL_EPSILON, and produces a boolean query otherwise. * tests/api_anydb.cc: Add tests for OP_MULT_WEIGHT operator with various queries, and various multipliers. * NEWS: Update Sun Sep 30 23:13:29 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/testsuite.h: Add TEST_NOT_EQUAL_DOUBLE() macro. * tests/harness/testsuite.cc: Add special case to TEST_EQUAL_DOUBLE_() for exact equality of the supplied arguments. This is tidier in general, because it avoids "inf" being produces by the calculation, but is only actually necessary in the case where the arguments are both exactly 0; it all works out nicely for other values of the arguments. Sun Sep 30 21:12:40 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/queryparsertest.cc: #define XAPIAN_DEPRECATED to disable compiler warnings about the deprecated forms of add_prefix() and add_boolean_prefix(). Sun Sep 30 21:07:25 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/Makefile.mk,docs/Makefile.am,docs/deprecation.rst, include/Makefile.mk,include/xapian.h,include/xapian/queryparser.h, queryparser/queryparser.cc,queryparser/queryparser.lemony, queryparser/queryparser_internal.h,tests/: Re-apply changes which needed more thought or more work before being included in a release. * NEWS: Update with current descriptions of these changes. Sun Sep 30 20:22:25 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/valgrind.supp: Add further suppressions needed for the version of zlib on ubuntu gutsy. Fri Sep 28 15:38:18 BST 2007 Olly Betts <olly@survex.com> * NEWS: Final (?) update for 1.0.3. Fri Sep 28 15:37:22 BST 2007 Olly Betts <olly@survex.com> * PLATFORMS: Another small update. Fri Sep 28 15:30:17 BST 2007 Olly Betts <olly@survex.com> * HACKING: Add link to the 1.0.N tracker bug. Fri Sep 28 14:04:06 BST 2007 Olly Betts <olly@survex.com> * PLATFORMS: More updates. Fri Sep 28 13:45:51 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/database.h: Document that empty metadata keys will cause an exception from get_metadata() and set_metadata(). * api/omdatabase.cc: Cause empty metadata keys to raise an InvalidArgumentError if they're passed to set_metadata() or get_metadata(). * tests/api_wrdb.cc: Add test_metadata3, to check behaviour with an empty metadata key. Fri Sep 28 13:30:49 BST 2007 Olly Betts <olly@survex.com> * PLATFORMS: Updates from tinderbox. Fri Sep 28 13:22:16 BST 2007 Olly Betts <olly@survex.com> * include/xapian/database.h: Clarify support for empty metadata keys. Fri Sep 28 11:00:08 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/valgrind.supp: Add second version of the zlib end condition check which works for the version of zlib in Ubuntu gutsy; ie, version 1.2.3.3, where the error occurs in the function called directly by deflate, rather than two levels below it. Fri Sep 28 03:11:11 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Update for 1.0.3. Fri Sep 28 03:05:10 BST 2007 Olly Betts <olly@survex.com> * api/Makefile.mk,docs/Makefile.am,docs/deprecation.rst, include/Makefile.mk,include/xapian.h,include/xapian/queryparser.h, queryparser/queryparser.cc,queryparser/queryparser.lemony, queryparser/queryparser_internal.h,tests/: Back out changes which need more thought or more work in the interests of getting 1.0.3 out this month. Fri Sep 28 03:04:19 BST 2007 Olly Betts <olly@survex.com> * NEWS: Update with recent changes. Fri Sep 28 01:55:42 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Revert to the previous deprecation policy. Fri Sep 28 01:20:34 BST 2007 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Rephrase since "should be in the MSet" might be misunderstood. Fri Sep 28 01:12:47 BST 2007 Olly Betts <olly@survex.com> * AUTHORS,docs/Makefile.am,docs/glossary.rst,docs/index.html: Add glossary put together by Jenny Black, also incorporating entries from Deron Meranda's glossary on the wiki. Thu Sep 27 23:06:30 BST 2007 Olly Betts <olly@survex.com> * docs/stemming.html: Reorder the initial paragraphs so we actually answer the question "What is a stemming algorithm?" up front. Thu Sep 27 22:15:36 BST 2007 Olly Betts <olly@survex.com> * matcher/valuerangepostlist.cc: Call ValueRangePostList::next() explicitly to avoid the overhead of virtual method dispatch. Thu Sep 27 14:43:25 BST 2007 Olly Betts <olly@survex.com> * common/termlist.h: Don't need to forward declare Xapian::Internal::ExpandWeight here. Thu Sep 27 14:36:00 BST 2007 Olly Betts <olly@survex.com> * common/termlist.h: Add missing (but indirectly included) '#include <xapian/base.h>'. Fri Sep 21 18:00:34 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Add deprecation for Database::positionlist_begin() throwing DocNotFoundError , for same reason as deprecating it throwing RangeError. Also, add deprecation for QueryParser::add_prefix(f, p) and add_boolean_prefix() - replaced by three argument form of add_prefix. Also, add a policy that we don't add XAPIAN_DEPRECATED when there's no replacement for the deprecated feature in the last x.x.0 release. * include/xapian/queryparser.h: Add @deprecated to documentation comments for add_prefix(f, p) and add_boolean_prefix() (but don't add XAPIAN_DEPRECATED() macro for now). Fri Sep 21 16:13:25 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/queryparser.h: Fix copy and paste errors in documentation comment, and hopefully clarify it somewhat. Fri Sep 21 15:50:40 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/queryparsertest.cc: Initialise the default prefix from prefixes.find(""), allowing multiple default prefixes to be set by calling add_prefix() with an empty field name. * queryparser/queryparser.lemony: Several more tests; mainly for the new default_prefix functionality, but also a bit more coverage for multiple prefixes in general. Fri Sep 21 15:04:02 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/queryparser.h,queryparser/queryparser.cc: API addition: Revert previous change to behaviour of QueryParser::add_prefix and QueryParser::add_boolean_prefix methods, to avoid API change within a release series. Add new QueryParser::add_prefix() form which takes three parameters; the third parameter is a value from a new "prefix_type" enum. * queryparser/queryparser.lemony,queryparser/queryparser_internal.h: Internal changes to allow multiple filters to be set for a single field: - Remove the PrefixInfoList object again - instead add a list of prefixes to PrefixInfo. We only aim to support one type of prefix for a given field, so the extra flexibility given by PrefixInfoList is a districation. - Change filter_group_id to be based on a list of prefixes, instead of a single prefix. - Change Term objects to store a list of prefixes instead of a single prefix. make_term() now takes a prefix argument and makes a single term. Term::as_... methods iterate through the prefix list and OR together the resulting terms. - Change TermList object to store a list of Terms instead of Queries, and also to keep track of whether the prefix lists for those Terms are all the same. When they are the same, generate separate phrases for each prefix, and OR them togeher, instead of generating phrases of "OR" groups of terms. Remove the unused "TermList::add_term()" method. * tests/queryparsertest.cc: Add simple tests of repeated inline fields, phrases with multiple prefixed fields, tests of backwards compatible behaviour of add_prefix(field, prefix), and add_boolean_prefix(), and of the new add_prefix() form with incompatible types. Thu Sep 20 02:13:36 BST 2007 Olly Betts <olly@survex.com> * NEWS: Update with changes since 1.0.2. Wed Sep 19 20:00:43 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Fix "Key_" in exception message to say "Key" (the result of overzealous search-and-replace). Wed Sep 19 17:08:16 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/queryparser.h,queryparser/queryparser.cc, queryparser/queryparser.lemony,queryparser/queryparser_internal.h: Handle repeated calls to QueryParser::add_boolean_prefix() with the same field - this will generate multiple terms for each occurrence of the field in the query. Involves reorganising the internal representation of prefixes, to hold a list of PrefixInfo objects for each field, instead of a single object. This also paves the way towards handling for multiple calls to QueryParser:add_prefix() with the same field, and allowing add_{boolean_}prefix() with an empty field to set the default prefix handling, but the behaviour of this are currently unchanged. Technically, this could be considered an API change, but the previous behaviour (of using only the most recent value set by add_boolean_prefix) was unintentional and undocumented, and seems unlikely to have been deliberately used. * tests/queryparsertest.cc: Add a very simple test for repeated boolean_prefix fields; several more are needed. Also, a commented out test of repeated non-boolean prefix fields. Tue Sep 18 23:28:48 BST 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.lemony: Remove unused variable "term_group". Tue Sep 18 16:56:37 BST 2007 Olly Betts <olly@survex.com> * api/matchspy.cc,include/xapian/matchspy.h: Use a set rather than a map to store the "multivalues" flags. Mon Sep 17 14:10:47 BST 2007 Olly Betts <olly@survex.com> * include/xapian/matchspy.h: Fix documentation comment typo. Sun Sep 16 04:13:20 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Fix compiler warning in mingw build. Sun Sep 16 02:59:54 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,tests/api_wrdb.cc: Check length of new terms is at most 245 bytes for flint in add_document() and replace_document() so that the API user gets an error there rather than when flush() is called (explicitly or implicitly). Fixes bug#44. Sat Sep 15 20:11:36 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_version.cc: Add missing include of msvc_posix_wrapper.h. Sat Sep 15 17:58:35 BST 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/database.cc, backends/flint/flint_database.cc,backends/flint/flint_database.h, backends/flint/flint_version.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h,common/database.h, include/xapian/database.h,tests/api_wrdb.cc: Add support for user specified metadata (bug#143). Sat Sep 15 02:26:40 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.cc,backends/flint/flint_cursor.h: Implement FlintCursor::find_entry_ge() a little more efficiently (we could do better still by modifying FlintTable too). Sat Sep 15 00:56:54 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.cc: If we found the exact key, just copy it to current_key. Fri Sep 14 21:52:45 BST 2007 Olly Betts <olly@survex.com> * backends/flint/: Overhaul FlintAllDocsPostList. Fri Sep 14 19:20:56 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_alltermslist.cc: Rework FlintAllTermsList::skip_to() to use FlintCursor::find_entry_ge(). Fri Sep 14 17:45:06 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.h: Improve documentation comment. Fri Sep 14 17:28:17 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spellingwordslist.cc, backends/flint/flint_synonym.cc: Use FlintCursor::find_entry_ge(). * backends/flint/flint_synonym.h: Use FlintCursor::find_entry_lt(). * backends/flint/flint_alltermslist.h: Use FlintCursor::find_entry_lt() and skip any keys before "\x00\xff" to allow for extra metadata keys. Fri Sep 14 15:26:04 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.cc,backends/flint/flint_cursor.h: The reason why a deleted key is still visible to the cursor is that it is running over the previous revision of the tree. So we don't need to call find_entry() at all - we can just call next(). Also extend FlintCursor::del() to return whether the cursor is positioned (like FlintCursor::next() does). * backends/flint/flint_postlist.cc: Use the return value of FlintCursor::del(). Fri Sep 14 15:10:39 BST 2007 Olly Betts <olly@survex.com> * tests/Makefile.am: Reorder the testdata/flint-x.y.z files consistently. Fri Sep 14 12:48:04 BST 2007 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Add "tar-ustar" option to AM_INIT_AUTOMAKE. This is needed to avoid a failure of make dist due to filepaths of more than 99 characters (in the generated documentation: the relevant filepaths are for the NumberValueRangeProcessor class, and extend to 103 characters). Automake documentation implies that tar-v7 is the default format, which doesn't support filepaths longer than 99 characters portably; ustar allows 256 characters. The automake documentation says that the ustar format "is believed to be old enough to be portable"; if we come across problems due to the format change we could consider renaming files to reduce the filepath length. * tests/Makefile.am: Remove testdata/flint-1.0.2/{value,position}.{baseA,DB} from distribution, since they don't actually exist. Fixes make dist. Fri Sep 14 10:45:35 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_cursor.cc,backends/flint/flint_cursor.h: Add FlintCursor::find_entry_gt(), and use it instead of flint_entry_ge() in FlintCuror::del(). This fixes failure of deldoc4 test. I'm not sure why find_entry() can return true immediately after the key it's searching for was deleted with B->del(), though: that merits further investigation. Fri Sep 14 04:57:14 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-inspect.cc: Avoid comparing char with 0 - on platforms where char is unsigned by default this can give a compiler warning. Fri Sep 14 03:40:18 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Removed unused variable. Fri Sep 14 01:56:36 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_alltermslist.cc: Fix comment typo. Fri Sep 14 01:51:52 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.cc,backends/flint/flint_cursor.h: Add FlintCursor::find_entry_lt() and FlintCursor::find_entry_ge() which we can implement more efficiently (in the future) than user code calling FlintCursor::find_entry() and then having to adjust the cursor position in some cases. Thu Sep 13 14:52:34 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_alltermslist.cc: FlintCursor::after_end() is never true after FlintCursor::find_entry() so remove check for this case! Thu Sep 13 03:11:19 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_alltermslist.h,backends/flint/flint_database.cc, backends/flint/flint_database.h: Don't pass FlintPostListTable pointer to FlintAllTermsList - it can find this via the database if we make it FlintDatabase instead of Database::Internal. Thu Sep 13 02:07:19 BST 2007 Olly Betts <olly@survex.com> * backends/flint/: Read/write the metainfo key from FlintDatabase to avoid having to pass a huge long list of values across once we start to store more statistics. Thu Sep 13 02:03:46 BST 2007 Olly Betts <olly@survex.com> * backends/database.cc,common/database.h: Move definitions of virtual methods out of the header file. Thu Sep 13 01:33:49 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_database.h: Store the total_length and lastdocid values in FlintDatabase object. Wed Sep 12 20:27:49 BST 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Eliminate a block which used to scope some variables which have now moved. Unify debug output from different cases after the match. Remove half-finished comment which doesn't seem to be trying to say anything useful. Wed Sep 12 20:20:18 BST 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc: In checkatleast2, the total number of matching documents in the database is 5, so use TEST_EQUAL instead of TEST_GREATER_OR_EQUAL to compare get_matches_lower_bound() with 5. Wed Sep 12 17:55:00 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.h: Fix some comment typos and outdated information. Remove some superfluous uses of "virtual". Wed Sep 12 17:15:58 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/quartz/quartz_database.cc: Failing to create a flint or quartz database because we couldn't create the directory for it now throws DatabaseCreateError not DatabaseOpeningError. * tests/api_db.cc: Update test cases. Wed Sep 12 15:40:16 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_database.h: If we reach the flush threshold during a transaction, flush the postlist changes, but don't actually commit them. Wed Sep 12 14:12:52 BST 2007 Olly Betts <olly@survex.com> * common/msvc_posix_wrapper.cc,common/msvc_posix_wrapper.h: Add msvc_posix_rename() which can rename a file on top of another file. * common/stringutils.h: Add common_prefix_length() function. * backends/flint/: Clean up FlintWritableDatabase - it now just inherits from FlintDatabase which allows several virtual methods which just forwarded to FlintDatabase to be dropped. Also, we now no longer need to pass FlintTable objects to other classes - they can just find the tables they want via the database pointer. The never-used "store_termfreqs" flag has been dropped from the termlist table entries - existing 1.0.x flint databases will be automatically upgraded to the new version. Opening a database now calls stat() less, so should be slightly more efficient. And TermIterator::positionlist_count() now works for the flint backend. * tests/Makefile.am,tests/api_db.cc,tests/testdata/flint-1.0.2/: New test flintbackwardcompat2 which tests that we can open a flint database from 1.0.2. * tests/api_wrdb.cc: New test adddoc4 which checks that termlists handle an initial term of any valid length correctly. * tests/testdata/flint-1.0.1/postlist.DB: Mark as a binary file in SVN. Tue Sep 11 23:02:56 BST 2007 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h: Implement TermIterator::positionlist_count() for the inmemory backend. * tests/api_posdb.cc: Add feature test for TermIterator::positionlist_count() (which currently skips for all backends except inmemory). Tue Sep 11 03:43:30 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Make sure flush_threshold gets initialised. * backends/flint/flint_database.h: Change type of flush_threshold from `size_t' to `Xapian::doccount' for consistency with the type of changes_made. Tue Sep 11 03:19:37 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_database.cc,backends/flint/flint_database.h: Change flush_threshold from static to non-static, and set it for each FlintWritableDatabase based on the value of the XAPIAN_FLUSH_THRESHOLD environment variable at the time the database was opened, instead of caching the value for each future database which is opened. Tue Sep 11 02:59:45 BST 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: When running rst2html, use "--exit-status=warning" rather than "--strict". The former actually gives a non-zero exit status for a warning or worse, while the former doesn't, but does include any "info" messages in the output HTML. Tue Sep 11 02:38:30 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Add "Database::positionlist_begin() throwing RangeError". Tue Sep 11 00:52:27 BST 2007 Olly Betts <olly@survex.com> * HACKING: Document a few more "coding standards". Mon Sep 10 21:34:23 BST 2007 Olly Betts <olly@survex.com> * examples/delve.cc,examples/simpleexpand.cc,net/tcpserver.cc, queryparser/queryparser.cc,queryparser/queryparser.lemony: Add more missing "#include <string.h>" instances. Mon Sep 10 20:24:25 BST 2007 Olly Betts <olly@survex.com> * net/serialise.cc: Add missing "#include <string.h>". Sun Sep 09 15:38:11 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.h,backends/flint/flint_synonym.h, backends/flint/flint_table.h: Fix typo (Z_DEFAULT_COMPRESSION where it should be Z_DEFAULT_STRATEGY) which meant that zlib compression wasn't enabled for the spelling or synonym tables. Fri Sep 07 19:31:44 BST 2007 Olly Betts <olly@survex.com> * docs/spelling.rst: Fix typo. Thu Sep 06 15:02:43 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/editdistance.cc: Fix off-by-one error in loop bounds when setting up fkp array initially. Fixes bug #194. * tests/api_wrdb.cc: Add regression test for #194, spell5. Thu Sep 06 14:31:15 BST 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: More localised fix for preserving the start of the valgrind report when in verbose mode. Rework the code which pulls out the start of the report to summarise why the test failed. Thu Sep 06 14:17:44 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/testsuite.cc: When displaying valgrind output, seek back to the start of the valgrind output for the current test before dumping it to stdout: previously, the earlier check for the valgrind error message to display (when not using -v) caused the first 1024 bytes of the valgrind output to get dropped. Tue Sep 04 20:13:29 BST 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: For remote submatches, keep a count of matches which we know exist, but we don't get passed from the submatch. These can be detected by checking for the lowerbound on the matches in the submatch being greater than the index of the last item in the returned submset. When the match is over, use this count, together with docs_matched, to adjust the lower bound and estimate (and upper bound, if we've not seen as many documents as we were asked to check for). Fixes checkatleast2 and checkatleast3 in the remote database case. * tests/harness/testsuite.h: Add TEST_GREATER_OR_EQUAL, TEST_GREATER, TEST_LESSER_OR_EQUAL and TEST_LESSER macros, which display the values of the arguments if they fail. * tests/api_anydb.cc: Use TEST_GREATER_OR_EQUAL and TEST_EQUAL in checkatleast2 and checkatleast3, to make failures easier to debug. Tue Sep 04 18:16:49 BST 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Restructure to make the control flow clearer. Rewrap some comments which got wrapped to ~82 columns somehow. Tue Sep 04 17:21:49 BST 2007 Olly Betts <olly@survex.com> * docs/overview.html: Restore the HTML header I accidentally deleted in November 2006. Tue Sep 04 17:15:30 BST 2007 Olly Betts <olly@survex.com> * docs/overview.html: Fix typo. Tue Sep 04 13:47:27 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_anydb.cc: Tighten test_checkatleast3 to check for previous bug (now fixed). Tue Sep 04 13:37:01 BST 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/multimatch.cc: If check_at_least is greater than maxitems, but there are fewer results than check_at_least, the lower and upper bounds reported should be equal (and exact). Fix a bug which caused this not to be the case, by checking if docs_matched < check_at_least, and forcing the bounds (and estimate) to be docs_matched in this case. Also, fix a bug when sorting by anything other than relevance, which was causing potential matches which are too low in the ranking to make the mset not to be added to the count of docs_matched, even when docs_matched < check_at_least. Also, add a few extra explanatory comments and debug log messages in get_mset(). Tue Sep 04 12:43:03 BST 2007 Olly Betts <olly@survex.com> * PLATFORMS: Update from debian buildd logs. Tue Sep 04 09:57:25 BST 2007 Richard Boulton <richard@lemurconsulting.com> * Makefile.am,configure.ac,msvc/genversion.pl,msvc/version.h.in: Move msvc/genversion.pl and msvc/version.h.in to win32msvc in xapian-maintainer-tools. Remove generation of msvc/version.h from configure.in, and AC_SUBST() of MAJOR_VERSION, MINOR_VERSION and REVISION, which were only in place for this generation. Remove msvc/version.h from distribution tarballs. Tue Sep 04 02:56:00 BST 2007 Olly Betts <olly@survex.com> * matcher/andpostlist.cc: If the database has > 2G documents then the sum of termfreq_min can overflow, so handle this case correctly. Tue Sep 04 00:05:47 BST 2007 Olly Betts <olly@survex.com> * xapian-config.in: We always need to include dependency_libs in the output of `xapian-config --libs` if shared libraries are disabled. Mon Sep 03 16:28:33 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Allow "xapian-check db/record." and "xapian-check db/record.DB". Mon Sep 03 02:16:43 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst,include/xapian/enquire.h: Deprecate Enquire::register_match_decider(). * api/omenquire.cc,common/omenquireinternal.h: Remove mdecider_map since the set values are never used anywhere. Sun Sep 02 20:54:33 BST 2007 Olly Betts <olly@survex.com> * HACKING: Add "update the 1.0.N tracker bug" to the release checklist. Fri Aug 31 10:44:19 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_db.cc,tests/api_wrdb.cc: Fix compilation of apitest, which was broken on 64 bit platforms due to size_t instead of doccount bug. Fri Aug 24 11:56:18 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/matchspy.cc,include/xapian/matchspy.h: Fix uses of size_t which should have been Xapian::doccount. Was causing compile errors on platforms where size_t and Xapian::doccount were different sizes. Fri Aug 03 20:10:15 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_nodb.cc: Add test_stringlistserialise1() to test the StringListSerialiser and StringListUnserialiser. Change the list of testcases to use the TESTCASE and END_OF_TESTCASES macros. Fri Aug 03 19:53:26 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/matchspy.cc,include/xapian/matchspy.h: Add StringListSerialiser and StringListUnserialiser classes which can be used to serialise a list of strings into a single string. Add flag to ValueCountMatchSpy to use this to unserialise the values read from the database into multiple strings, allowing several values to be stored for a particular slot in a database: this allows multiple values of a particular facet to be stored. Thu Jul 26 16:55:18 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_table.cc: Change STRINGIZE to om_tostring() in error message about a key which is too long; STRINGIZE doesn't work since FLINT_BTREE_MAX_KEY_LEN is no longer a preprocessor constant - it's a C++ constant; before this change, the error message contained the literal text "FLINT_BTREE_MAX_KEY_LEN". Tue Jul 17 12:20:55 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/matchspy.cc,include/xapian/matchspy.h: Add a MultipleMatchDecider, which calls a list of deciders in order, until one returns false. * include/xapian/enquire.h: Expand comment to say which way round the return values of the MatchDecider operator are. * tests/api_db.cc: Add test_matchfunctor3() to test the MultipleMatchDecider - also tests the ValueCountMatchSpy, incidentally. Tue Jul 17 11:10:42 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/matchspy.cc,include/xapian/matchspy.h: Make get_most_frequent_items() private, since I'm not convinced it's useful to make it available externally. We can always make it public again if it's useful to do so. * test/api_wrdb.cc: Update tests accordingly. Sat Jul 14 00:17:06 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/matchspy.cc,include/xapian/matchspy.h: Get rid of the TopValueMatchSpy in favour of a new function "get_most_frequent_items", which processes the output of a ValueCountMatchSpy. Add a TermCountMatchSpy which counts the occurrences of terms with a given prefix, and gives output in the same form as ValueCountMatchSpy. Add convenience methods ValueCountMatchSpy::get_top_values() and TermCountMatchSpy::get_top_terms() which use this to return the most frequent items seen by the matchspy. * tests/api_wrdb.cc: Test the ValueCountMatchSpy, and the get_most_frequent_items() function. Fri Jul 13 18:57:30 BST 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/andpostlist.cc,tests/api_anydb.cc: Improve the lower bound on the number of matching documents for an AND query - if the sum of the lower bounds for the two sides is greater than the number of documents in the database, then some of them must have both terms. This greatly improves the lower bound for queries of the form "<alldocuments> FILTER foo" (since FILTER is a subclass of AND, and uses the same method), but is probably a useful improvement elsewhere. Fri Jul 13 17:53:46 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/valueranges.rst: Correct out-of-date reference to float_to_string, spotted by Enrico Zini. Wed Jul 11 22:54:43 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/matchspy.cc,include/xapian/matchspy.h,tests/api_wrdb.cc: Move ValueAndFrequency out of the TopValueMatchSpy class - it doesn't seem to be possible to wrap it with swig without doing this, and it doesn't seem much less tidy. Wed Jul 11 18:03:18 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/matchspy.cc,include/xapian/matchspy.h: Split MatchSpy into a base class (ValueCountMatchSpy), which just counts the values in the documents it sees, and a subclass (CategorySelectMatchSpy) which does the grouping into categories, and scores the categories. Add another subclass (TopValueMatchSpy) which calculates a sorted vector holding the most frequent values. * tests/api_wrdb.cc: Update matchspy1 and matchspy2 to use CategorySelectMatchSpy instead of MatchSpy, and corresponding changes to the methods. Add matchspy3 to test the TopValueMatchSpy. Mon Jul 09 15:40:58 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/termgentest.cc: Fix copyright: everything I've done in 2007 has been for Lemur. Mon Jul 09 15:40:38 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/version_h.cc: Fix type in comment. Fri Jul 06 12:14:48 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/enquire.h: Add note to register_match_decider() documentation comment of what this function is for, and also that it doesn't currently do anything (other than store the values passed to it in the enquire internals). Thu Jul 05 20:52:52 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/categorisation.rst: Fix typo. Thu Jul 05 20:41:49 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_wrdb.cc: Rework matchspy2 to use approximate values to avoid failing on some platform due to differing precision. Thu Jul 05 12:21:48 BST 2007 Olly Betts <olly@survex.com> * docs/categorisation.rst: Complete half sentence. Thu Jul 05 12:17:20 BST 2007 Olly Betts <olly@survex.com> * docs/categorisation.rst: Add section on restricting by category values. Thu Jul 05 04:47:18 BST 2007 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/categorisation.rst: Add topic document discussing use of Xapian::MatchSpy. Thu Jul 05 02:08:13 BST 2007 Olly Betts <olly@survex.com> * include/xapian.h,tests/api_wrdb.cc: Restore MatchSpy stuff. Thu Jul 05 01:38:38 BST 2007 Olly Betts <olly@survex.com> * HACKING: Note that RoadMap on the wiki needs updating for each release. Thu Jul 05 01:03:05 BST 2007 Olly Betts <olly@survex.com> * AUTHORS: Update. Thu Jul 05 00:31:00 BST 2007 Olly Betts <olly@survex.com> * NEWS: Minor tweaks and clarifications. Thu Jul 05 00:29:53 BST 2007 Olly Betts <olly@survex.com> * include/xapian.h,tests/api_wrdb.cc: Disable MatchSpy class for 1.0.2 as it's not ready for release yet. Wed Jul 04 21:20:59 BST 2007 Richard Boulton <richard@lemurconsulting.com> * NEWS: Update with release date for release 1.0.2 Wed Jul 04 21:09:18 BST 2007 Richard Boulton <richard@lemurconsulting.com> * AUTHORS: Add Tomek Jackowiak, for reporting bug #166 (missing Xapian::Query documentation). Wed Jul 04 20:45:56 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Updated version and library version for 1.0.2. Wed Jul 04 20:36:07 BST 2007 Olly Betts <olly@survex.com> * api/matchspy.cc,include/xapian/matchspy.h,tests/api_wrdb.cc: Tidy up API for MatchSpy::build_numeric_ranges() and add tests. Wed Jul 04 19:40:31 BST 2007 Richard Boulton <richard@lemurconsulting.com> * NEWS: Fully updated. Wed Jul 04 18:33:59 BST 2007 Richard Boulton <richard@lemurconsulting.com> * NEWS: Move some of the changelog entries into news-style comments. Wed Jul 04 17:50:19 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_wrdb.cc: SKIP_TEST() spell3 on windows - it doesn't work there due to trying to delete (and then overwrite) an open database. We shouldn't be doing this, so add a long FIXME explaining what we should be doing instead. Wed Jul 04 17:27:22 BST 2007 Olly Betts <olly@survex.com> * NEWS: Partly update. Wed Jul 04 17:19:51 BST 2007 Olly Betts <olly@survex.com> * AUTHORS: Add Simon Tatham for spotting the -INFINITY issue, and for a lot of useful advice during the CVS to SVN transition. Wed Jul 04 17:16:59 BST 2007 Olly Betts <olly@survex.com> * api/,include/xapian/queryparser.h,tests/queryparsertest.cc: Rename Xapian::NumberValueRangeProcessor::float_to_string() to Xapian::sortable_serialise() and Xapian::NumberValueRangeProcessor::string_to_float() to Xapian::sortable_unserialise(). Wed Jul 04 17:16:16 BST 2007 Olly Betts <olly@survex.com> * HACKING: Add rationale for 2 space indents for "public", etc. Wed Jul 04 14:09:57 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Loops on integers and scale to doubles rather than looping on doubles and scaling to get integers. Wed Jul 04 13:01:11 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: If the NumberValueRangeProcessor string encodings don't sort the same way, there's no need to report the string encodings in the error message as they are written to "tout" just before, along with the respective numbers they came from. The string encodings can contain control characters, so are liable to corrupt the test failure message. Wed Jul 04 12:25:15 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/valuerangeproc.cc: Fix encoding of extremely large, negative, non-IEEE representation to be -INFINITY, not +INFINITY. Thanks to Simon Tatham for pointing this out. * include/xapian/queryparser.h: Add a documentation comment note that zero and -zero will be transformed to the same value. Wed Jul 04 10:09:33 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/valuerangeproc.cc: Change encoding used for doubles to a much more compact scheme, and add support for +-INFINITY. Also, some other tweaks to avoid unnecessary computation. * tests/queryparsertest.cc: Update test cases to expect new encoding of doubles, and add values for INFINITY and DBL_MAX and some other special numbers to the list of numbers that are specifically checked. Wed Jul 04 03:29:10 BST 2007 Olly Betts <olly@survex.com> * api/matchspy.cc,include/xapian/matchspy.h,tests/api_wrdb.cc: Category score now explicitly weights for having nearer the requested number of categories. Category score is now normalised so values should be roughly comparable between different sized collections. Add untested code to build ranges. Wed Jul 04 00:48:07 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Add new "--no-renumber" option to preserve document ids from source databases. Fix bug in change to add support for spelling and synonym tables which was preventing any renumbering from happening! Wed Jul 04 00:25:52 BST 2007 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h,tests/api_wrdb.cc: Fix bug in inmemory backend - using replace_document() to add a document with a specific document id above the highest currently used would create empty documents for all document ids in between. Tue Jul 03 09:27:36 BST 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Add note that protection level declarations in classes and structs should only be indented by 2 spaces. Tue Jul 03 03:53:11 BST 2007 Olly Betts <olly@survex.com> * api/Makefile.mk,api/matchspy.cc,include/xapian/matchspy.h, tests/api_wrdb.cc: Add MatchSpy method to score how good a categorisation is. Tue Jul 03 01:24:19 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Replace `#include <xapian/stem.h>' with forward declaration of Stem. Tue Jul 03 01:14:12 BST 2007 Olly Betts <olly@survex.com> * include/xapian/errorhandler.h: Replace `#include <xapian/error.h>' with forward declaration of Error. * api/errorhandler.cc: Now needs to explicitly `#include <xapian/error.h>'. Tue Jul 03 01:08:30 BST 2007 Olly Betts <olly@survex.com> * include/xapian/dbfactory.h: Replace `#include <xapian/database.h>' with forward declarations of Database and WritableDatabase. Tue Jul 03 00:59:57 BST 2007 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Remove `#include <xapian/error.h>'. Mon Jul 02 23:25:55 BST 2007 Olly Betts <olly@survex.com> * include/Makefile.mk,include/xapian.h,include/xapian/matchspy.h, tests/api_wrdb.cc: Implement "MatchSpy" class to tally values in matching documents. Mon Jul 02 18:15:18 BST 2007 Olly Betts <olly@survex.com> * api/omenquire.cc,include/xapian/enquire.h: Fix last change to preserve ABI compatibility. Mon Jul 02 17:52:10 BST 2007 Olly Betts <olly@survex.com> * api/omenquire.cc,common/multimatch.h,common/omenquireinternal.h, include/xapian/enquire.h,matcher/multimatch.cc,net/remoteserver.cc, tests/api_db.cc: Add support for a "matchspy" - a MatchDecider which is documented to be tested on ever candidate document. We plan to optimise the current matchdecider to be used as little as possible. Mon Jul 02 17:35:12 BST 2007 Olly Betts <olly@survex.com> * tests/api_db.cc: Enhance matchfunctor1 to test that the functor returns ALL matching documents, not just that all documents returned match. Mon Jul 02 16:36:57 BST 2007 Richard Boulton <richard@lemurconsulting.com> * common/omassert.h: Fix to avoid warning in gcc-snapshot; use do{}while(0) to protect assertions, and remove the old "if (a) {} else {fail}" constructions which tried to do the same job. Mon Jul 02 14:56:58 BST 2007 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-inspect.cc: Add missing #include of <stdio.h> - fixes gcc-2.95 compilation. Mon Jul 02 14:54:50 BST 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/termgenerator_internal.h: Fix for gcc-2.95 compilation. Mon Jul 02 14:09:22 BST 2007 Olly Betts <olly@survex.com> * api/Makefile.mk,api/valuerangeproccompat.cc, include/xapian/queryparser.h: Put the new NumberValueRangeProcessor in a sub-namespace which we then import it from. Restore the old NumberValueRangeProcessor implementation inside the library so that it is available to code linked against 1.0.0 or 1.0.1 to keep ABI compatibility. Sun Jul 01 15:32:30 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Clearer description of NumberValueRangeProcessor. Don't document the number format in terms of strtod() - that's a bug, not a feature we want to be tied to. * include/xapian/query.h,include/xapian/queryparser.h: Use "@a" not "\a" for consistency with existing usage. Sun Jul 01 10:09:42 BST 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/localmatch.cc: When registering a term, check if it's empty, and if so use the document count as the term frequency, rather than calling the database to get the term frequency of an empty term (most database backends have an assertion that the term is not empty, and an empty term isn't a valid parameter to pass to get_termfreq() anyway). * matcher/multimatch.cc: For MatchAll queries, the "" term will not be found in the termlist of the top document, so check if a "" term is present in the query, and use it to increase the percent_scale if so. This fixes an Assertion which checked that percent_scale was not 0. * tests/api_db.cc: Add test_matchall1() which tests running a query with a MatchAll query. * tests/api_nodb.cc: Rename test_emptyquery1 and test_emptyquery2 to test_emptyquery2 and test_emptyquery3, respectively. There was already a (different) test_emptyquery1 in api_anydb.cc. Sat Jun 30 18:04:06 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Write `2.0' rather than `(double)2'. Sat Jun 30 16:12:49 BST 2007 Olly Betts <olly@survex.com> * docs/valueranges.rst: Fix typo. Fix ".. note:" to ".. note::" so it appears visibly in the HTML output rather than in an HTML comment! Sat Jun 30 15:14:33 BST 2007 Olly Betts <olly@survex.com> * api/omdocument.cc,common/document.h,include/xapian/document.h: Add Xapian::Document::get_docid() method. Sat Jun 30 04:00:57 BST 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: If there's only one term in the query (a pretty common case) we don't need to look at the top document's termlist to determine that it matches all the query terms. Sat Jun 30 00:37:09 BST 2007 Olly Betts <olly@survex.com> * INSTALL: Add note that zlib must be installed before you can build Xapian. Fri Jun 29 22:40:58 BST 2007 Olly Betts <olly@survex.com> * backends/flint/: If doccount == lastdocid, all document ids up to lastdocid are used, so we provide a special really efficient version implementation of iterating all documents for this common case for flint. Fri Jun 29 00:39:04 BST 2007 Olly Betts <olly@survex.com> * common/stringutils.h: Add startswith() and endswith() overloads which take a single character. * backends/flint/flint_spellingwordslist.cc,bin/xapian-inspect.cc: Use new forms. * backends/flint/flint_spellingwordslist.cc: Fix comment typo. Thu Jun 28 21:43:49 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spellingwordslist.cc: Another begins_with() which I somehow missed. Thu Jun 28 19:51:03 BST 2007 Olly Betts <olly@survex.com> * api/valuerangeproc.cc,backends/flint/flint_alltermslist.cc, backends/flint/flint_synonym.cc, backends/inmemory/inmemory_alltermslist.cc, backends/quartz/quartz_alltermslist.cc, backends/quartz/quartz_metafile.cc,bin/xapian-inspect.cc, common/stringutils.h,tests/harness/testsuite.cc: Rename begins_with() to startswith() and ends_with() to endswith() for consistency with Python string operations. Thu Jun 28 18:49:50 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Don't put a Z prefix on terms if the stemmer is "none", which matches what TermGenerator generates. Thu Jun 28 02:22:04 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Don't SEGV when Query::MatchNothing is used with OP_AND_NOT (fixes bug#176). * tests/api_nodb.cc: Add regression test emptyquery2. * AUTHORS: Thank bug reporter (Krzysztof Klemm). Wed Jun 27 13:05:55 BST 2007 Richard Boulton <richard@lemurconsulting.com> * NEWS: Updated. Wed Jun 27 13:04:19 BST 2007 Richard Boulton <richard@lemurconsulting.com> * examples/copydatabase.cc: Add support for copying the spelling data. Wed Jun 27 12:46:54 BST 2007 Richard Boulton <richard@lemurconsulting.com> * NEWS: Updated. Wed Jun 27 12:13:50 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-check.cc: Check spelling and synonym Btrees. Wed Jun 27 12:10:47 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Add merging for spelling and synonym tables. Wed Jun 27 12:09:34 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.cc: Fix PrefixCompressedStringItor to not ignore the last entry. * tests/api_wrdb.cc: Regression test (spell4). Wed Jun 27 11:51:10 BST 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Remove bogus comment. Wed Jun 27 04:12:56 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.h: Make FlintCursor::after_end() const. Tue Jun 26 22:31:24 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_wrdb.cc: Add test_spell3(), which tests spelling correction with multi databases, and the results of the iterator from Database::spellings_begin() with single and multi databases. Tue Jun 26 20:51:13 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/omtermlistiterator.cc: Handle pruning, so that multi-database spelling word iterators work. Pruning should arguably always have been handled, but it's never been possible for the TermList passed to TermIterator to prune before, so it's not been an issue. Tue Jun 26 20:50:06 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_spellingwordslist.h: Don't bother testing whether we found an entry which was exactly "W" - if we do, the database is corrupt, but the best recovery strategy would just be to continue at the next matching entry. Tue Jun 26 20:26:03 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc: Remove mistaken skip_to("W") - should be done by open_spelling_wordlist() rather than here. * backends/flint/flint_spellingwordslist.h: Only call cursor->prev() if find_entry("W") finds an entry. Tue Jun 26 17:46:27 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_alltermslist.cc, backends/flint/flint_spellingwordslist.cc, backends/flint/flint_synonym.cc: Remove calls to abort() which are followed immediately by an Assert() which checks the same condition. Tue Jun 26 17:26:02 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/database.h: Add spellings_begin() and spellings_end(), to iterate through the spelling correction target words. * api/omdatabase.cc: Implement spellings_begin(), using the newly added FreqAdderOrTermList to join the spellings lists. * common/database.h,backends/database.cc: Add open_spelling_wordlist() to get iterator of spelling targets, if any, with default implementation which returns NULL. * backends/flint/flint_database.h,backends/flint/flint_database.cc: Implement open_spelling_wordlist() for flint. * backends/flint/flint_spelling.h: Make merge_changes() public, so it can be called when a writable database needs to open a spelling wordlist. * backends/flint/flint_spellingwordslist.h, backends/flint/flint_spellingwordslist.cc: New files, implementing an iterator over the spelling targets. Tue Jun 26 17:25:09 BST 2007 Richard Boulton <richard@lemurconsulting.com> * common/ortermlist.h,expand/ortermlist.cc: Add a "FreqAdderOrTermList", which adds the term frequencies of the sublists, rather than requiring that they're equal for terms which are in both. Tue Jun 26 17:22:03 BST 2007 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Update to mention spelling and synonym tables, and to document optional tables more clearly. Tue Jun 26 16:23:25 BST 2007 Olly Betts <olly@survex.com> * tests/termgentest.cc: Add feature test for TermGenerator spelling integration. Tue Jun 26 16:14:56 BST 2007 Olly Betts <olly@survex.com> * tests/termgentest.cc: Fix comment paste-o. Tue Jun 26 16:11:11 BST 2007 Olly Betts <olly@survex.com> * include/xapian/termgenerator.h,queryparser/termgenerator.cc, queryparser/termgenerator_internal.cc, queryparser/termgenerator_internal.h: Add support for generating spelling data to TermGenerator class. Tue Jun 26 15:22:12 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Don't skip "empty" source databases, as they may have spelling and/or synonym data. Warn if more than one source database has spelling data and ignore all but the first. Similarly warn and ignore for synonym data. Tue Jun 26 13:52:15 BST 2007 Olly Betts <olly@survex.com> * examples/copydatabase.cc: Fix code which extracts the leafname to handle a trailing directory separator. * examples/copydatabase.cc: Copy synonym data across. Print warning that spelling data isn't copied (as there's no API to access it currently). Tue Jun 26 13:21:13 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Fix to compact spelling and synonym tables. Tue Jun 26 02:20:50 BST 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Fix randomly indented line. Tue Jun 26 02:20:01 BST 2007 Olly Betts <olly@survex.com> * include/xapian/document.h: Change parameter name for Document::get_value() to match other parameter names in the class. Tue Jun 26 02:19:16 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: FLAG_AUTO_MULTIWORD_SYNONYMS now implies FLAG_AUTO_SYNONYMS. Tue Jun 26 02:18:02 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add test for a single word synonym in multi-synonym mode. Tue Jun 26 02:08:47 BST 2007 Olly Betts <olly@survex.com> * common/ortermlist.h,queryparser/termgenerator_internal.h: Fix warnings from Intel's C++ compiler. Tue Jun 26 01:45:09 BST 2007 Olly Betts <olly@survex.com> * HACKING: Snapshots and releases are now bootstrapped with libtool 1.5.24, which includes all the patches we were applying on top of 1.5.22. * HACKING: Expand note about preferring pre-increment to post-increment to mention preferring it to adding one, and to cover decrementing too. Tue Jun 26 01:34:41 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.cc: Fix bug appending spelling data to an existing trigram. Tue Jun 26 01:34:07 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Fix handling of exceptions during commit. Mon Jun 25 17:01:40 BST 2007 Olly Betts <olly@survex.com> * docs/spelling.rst: Update for Unicode spelling correction. Mon Jun 25 15:41:43 BST 2007 Olly Betts <olly@survex.com> * api/editdistance.cc,api/editdistance.h,api/omdatabase.cc: Convert to UTF-32 before calculating edit distances. * tests/api_wrdb.cc: Add test coverage for Unicode spelling correction. Mon Jun 25 14:03:34 BST 2007 Olly Betts <olly@survex.com> * tests/quartztest.cc: Remove tests for removed method QuartzPostList::get_collection_freq(). Mon Jun 25 12:32:07 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.cc: Replace abort() with exception throwing. Mon Jun 25 03:01:20 BST 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Track the fd used to communicate with a xapian-tcpsrv child process so we can close it when we reap the child pid. Mon Jun 25 02:44:04 BST 2007 Olly Betts <olly@survex.com> * docs/valueranges.rst: Fix bad RST markup. Sun Jun 24 23:24:26 BST 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Fix memory leak in test harness launching of xapian-tcpsrv so that "apitest -b remotetcp" now passes under valgrind. Sun Jun 24 19:39:59 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/valueranges.rst: Add documentation for the NumberValueRangeProcessor, now that it is fixed. Sun Jun 24 14:29:06 BST 2007 Olly Betts <olly@survex.com> * backends/remote/net_postlist.cc,backends/remote/net_postlist.h: Move NetworkPostList virtual methods out of the header. Move the constructor into the header, since that can be inlined. Sun Jun 24 14:15:43 BST 2007 Olly Betts <olly@survex.com> * backends/flint/,backends/multi/multi_postlist.cc, backends/multi/multi_postlist.h,backends/quartz/, backends/remote/net_postlist.cc,backends/remote/net_postlist.h, backends/remote/remote-database.cc,common/emptypostlist.h, common/postlist.h,common/remote-database.h: For Flint and Quartz, refactor so we no longer create a postlist just to find the termfreq or collection frequency. These were the only places which used PostList::get_collection_freq(), so eliminate it (bug#124). Sun Jun 24 09:40:51 BST 2007 Richard Boulton <richard@lemurconsulting.com> * msvc/genversion.pl: Add a script to generate the version.h file for MSVC builds by parsing configure.ac to extract the version number, and then performing the appropriate substitutions on msvc/version.h.in. Shouldn't be used for tarball builds (because version.h already exists) or for builds which use configure. Sun Jun 24 02:25:38 BST 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: The check_at_least parameter to Enquire::get_mset() is now handled in a more efficient way - no extra memory is now required, and CPU overhead should be reduced (bug#174). Sun Jun 24 02:00:07 BST 2007 Olly Betts <olly@survex.com> * docs/synonyms.rst,docs/Makefile.am: New "topic" document describing synonym support. * docs/Makefile.am: HTML documentation generated from RST files wasn't being installed. Sun Jun 24 00:45:46 BST 2007 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remote-database.h, common/remoteprotocol.h,common/submatch.h,docs/remote_protocol.html, matcher/,net/remoteserver.cc: Pass check_at_least to the remote server to reduce the amount of work required to produce the match on the remote server, and also reduce the serialised size of the returned MSet. Sun Jun 24 00:17:18 BST 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc: More check_at_least tests. Sat Jun 23 20:02:48 BST 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc,tests/api_wrdb.cc: Add multi-database support for synonyms. Sat Jun 23 18:50:59 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/database.h: Tiny fix to a documentation comment. Sat Jun 23 17:35:37 BST 2007 Olly Betts <olly@survex.com> * common/ortermlist.h,expand/ortermlist.cc: Comment out unused method OrTermList::get_collection_freq(). The same dummy implementation is provided by the parent class (TermList). Sat Jun 23 17:07:40 BST 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/database.cc,backends/flint/, common/database.h,include/xapian/database.h,tests/api_wrdb.cc: Implement Database::synonym_keys_begin() to allow iteration over all the terms for which synonyms have been added. Sat Jun 23 04:19:54 BST 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix handling of check_at_least parameter - we weren't discarding matches above the requested MSet size correctly. * tests/api_anydb.cc: Add regression test checkatleast2. Sat Jun 23 03:00:45 BST 2007 Olly Betts <olly@survex.com> * docs/queryparser.html: Document synonyms. Sat Jun 23 02:50:55 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Unapply commented out fragments towards supporting '~' on phrases, which I didn't mean to apply. Sat Jun 23 02:25:32 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.lemony, tests/queryparsertest.cc: Add support for synonym operator ('~') to the QueryParser. Sat Jun 23 01:52:51 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix two bugs in the query parser - now '+' and '-' work on bracketed subexpressions as documented. * tests/queryparsertest.cc: Add regression tests. Sat Jun 23 01:47:30 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.cc: Use "using namespace std;". Fri Jun 22 18:48:21 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_synonym.cc: Fix iteration of unflushed synonyms. Fri Jun 22 18:21:39 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Fix typo. Fri Jun 22 18:02:21 BST 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/database.cc,backends/flint/, common/database.h,include/xapian/database.h, include/xapian/queryparser.h,queryparser/queryparser.lemony: Add support for storing synonyms for terms in a new optional Btree table in flint databases, and using them in the QueryParser. * tests/queryparsertest.cc: Feature tests for synonyms. Fri Jun 22 17:58:37 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Removed superfluous "private:" in class State. Make State::get_database() const. Fri Jun 22 17:58:15 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Make sure .flint subdirectory exists. Fri Jun 22 17:52:53 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-inspect.cc: Don't die with an error if the user tries to move forward twice from the end. Also, make "prev" from the end work. Fri Jun 22 17:47:36 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: If an input token has a yyminor, report its name in the debug output. Fri Jun 22 14:47:17 BST 2007 Olly Betts <olly@survex.com> * NEWS: Updated with changes since 1.0.1. Thu Jun 21 21:12:49 BST 2007 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remoteprotocol.h, common/remoteserver.h,net/remoteserver.cc: Protocol version increased to 30.2. This difference from 30.1 is that MSG_DELETEDOCUMENT now send REPLY_DONE to allow DocNotFoundError to be propagated. MSG_DELETEDOCUMENT now has a new number and the old number for MSG_DELETEDOCUMENT is now MSG_DELETEDOCUMENT_PRE_30_2 which is handled by the server so that older clients will continue to work. * docs/remote_protocol.html: Update. * tests/api_wrdb.cc: Add regression test for DocNotFoundError bug. Thu Jun 21 17:22:53 BST 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc,tests/queryparsertest.cc,docs/spelling.rst: Add support for spelling correction when using multiple databases. Thu Jun 21 16:15:22 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add feature tests for correcting spelling errors in the QueryParser. Thu Jun 21 15:01:00 BST 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Check the edit distance 3 words are found if asked for. Thu Jun 21 14:59:15 BST 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc,common/omdebug.h: Add some debug tracing for the spelling correction. Thu Jun 21 13:45:12 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.cc,docs/spelling.rst: Generate "bookends" for four character words too, so that we suggest "fuor" -> "four". * tests/api_wrdb.cc: More testcases. Thu Jun 21 13:08:07 BST 2007 Olly Betts <olly@survex.com> * docs/spelling.rst: Assorted minor improvements. Add note about not detecting single character substitutions in two character words. Thu Jun 21 13:02:57 BST 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Check cases for single edits to a two character word work as expected. Thu Jun 21 04:13:46 BST 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Add test cases to check we find all single edit errors for a three letter word. Thu Jun 21 04:08:11 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.h: Now that FlintSpellingTable::merge_changes() is only called from within the class, make it a private method. discard_changes() is only used once, and from within the class, so just inline it as that's clearer. Make the comment about the overridden methods a doxygen grouping comment, Thu Jun 21 04:02:51 BST 2007 Olly Betts <olly@survex.com> * docs/spelling.rst: Update to mention "bookend" bigrams. Thu Jun 21 03:56:52 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.cc: Generate "bookend" spelling entries for two and three letter terms, consisting of the prefix 'B' followed by the first and last letters. This allows us to handle substitution or deletion of the middle character of a three letter word, or insertion in the middle of a two letter word. Thu Jun 21 03:18:28 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Some doubles in the list which value_range_serialise1 tests have the same values on some platforms, so test that adjacent numbers and their string counterparts compare the same way rather than that both are "<". Thu Jun 21 02:46:01 BST 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Add "spell1" - a simple feature test for spelling based on Richard's python test. Use TESTCASE() and END_OF_TESTCASES macros. Wed Jun 20 21:57:01 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: WritableDatabase::delete_document() no longer cancels pending changes if the document doesn't exist. * backends/quartz/quartz_database.cc: Same fix. Wed Jun 20 20:53:52 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: We may now need to flush even if there have been no documents added/replaced/deleted, as there may be changes to the spelling table. Wed Jun 20 20:46:30 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Fix a function name in the debug loggging. Wed Jun 20 20:39:43 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.cc: Fix FlintSpellingTermList not to report at_end() one entry too early. Wed Jun 20 19:21:45 BST 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Don't touch <dbdir>/log for flint as flint doesn't create a log like quartz does. Wed Jun 20 17:10:29 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_spelling.h: Push special handling for spelling_table into the FlintSpellingTable class. This fixes "is_modified()" to take into account unflushed changes held in memory. Wed Jun 20 13:59:01 BST 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc: Don't use RETURN() in methods which return void. Wed Jun 20 11:56:30 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/editdistance.cc: Change an Assert in is_transposed() (which fails in some cases) to a test, since it seems legitimate that a value of pos2 <= 0 might be supplied to the function (though the answer will always be "false" in that case). Wed Jun 20 02:38:51 BST 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc: Fix two incorrect return types in debug logging. Wed Jun 20 02:14:40 BST 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc: Add missing implementations of WritableDatabase::add_spelling() and WritableDatabase::remove_spelling(). Wed Jun 20 00:58:46 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.h: Make the maps private. Wed Jun 20 00:05:15 BST 2007 Olly Betts <olly@survex.com> * HACKING: Snapshots now bootstrapped with automake 1.10. Wed Jun 20 00:03:42 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_spelling.cc,common/Makefile.mk, expand/Makefile.mk,expand/ortermlist.h: Move ortermlist.h from "expand/" to "common/" so we can included it from flint_spelling.cc without a relative path. Tue Jun 19 22:41:08 BST 2007 Olly Betts <olly@survex.com> * api/,backends/database.cc,backends/flint/,common/database.h, docs/Makefile.am,docs/spelling.rst,include/xapian/database.h, include/xapian/queryparser.h,queryparser/queryparser.cc, queryparser/queryparser.lemony,queryparser/queryparser_internal.h: Add support for spelling correction. Tue Jun 19 19:18:40 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_table.cc, backends/flint/flint_table.h: Add new method FlintTable::key_exists() so we don't need to create a cursor to check if a key exists in FlintDatabase::term_exists() (the other alternative previously was to read the key and tag and ignore the tag, but that could involve reading more blocks if the tag is long, and it could involve uncompressing the tag too). Tue Jun 19 17:49:09 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/queryparsertest.cc: Cast the input to pow() to a double, to fix a problem on windows (which has several versions of pow(), and the version taking a float was being used by default, resulting in some of the tests overflowing the range of float). Tue Jun 19 17:48:42 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/valuerangeproc.cc: Fix assertions, so that the code works with assertions turned on. Tue Jun 19 12:35:46 BST 2007 Olly Betts <olly@survex.com> * matcher/branchpostlist.cc: Add missing #include <config.h>. * matcher/branchpostlist.cc,matcher/branchpostlist.h: Move other virtual method definitions out of the header. Mon Jun 18 21:41:15 BST 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/branchpostlist.h: Add missed copyright statement. Mon Jun 18 21:36:31 BST 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/branchpostlist.cc,matcher/branchpostlist.h: Add get_wdf() method to BranchPostList (which just sums the wdf, which is useful for a synonym postlist, which is the only situation in which this will be called). Add the method definition in a separate file, as per the comment I just added to HACKING. * matcher/Makefile.mk: Add branchpostlist.cc Mon Jun 18 21:33:56 BST 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Add note about virtual methods (saying that the shouldn't be defined in header files). Mon Jun 18 20:15:18 BST 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/phrasepostlist.cc,matcher/phrasepostlist.h, matcher/exactphrasepostlist.cc,matcher/exactphrasepostlist.cc: Add get_wdf() methods to NearPostList, PhrasePostList and ExactPhrasePostList - these use a very rough approximation (but there's a big long comment justifying it). Mon Jun 18 18:38:43 BST 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/mergepostlist.cc,matcher/mergepostlist.h: Add copyright statements for previous commit (and update license while I'm at it). Mon Jun 18 18:24:43 BST 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/mergepostlist.cc,matcher/mergepostlist.h: Add get_wdf() method to MergePostList (just passes through the value from the current postlist). Not currently needed, but the expected behaviour is obvious, and this will be needed if there's a synonym postlist higher in the query tree (when synonym postlists are implemented). Also, correct a documentation comment typo. Mon Jun 18 13:09:36 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/valuerangeproc.cc: Cast the first parameter to ldexp() to double to fix compilation on some platforms, and add a few more explanatory code comments to NumberValueRangeProcessor::string_to_float() Mon Jun 18 11:33:49 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/valuerangeproc.cc: Add static methods to NumberValueRangeProcessor to serialise and unserialise doubles. Change the code which tests a term for being a number to use strtod, and accept any string which strtod thinks is a number. If NumberValueRangeProcessor accepts a range, it now converts the begin and end values to strings using the serialise method. Also, expand some of the code comments. * include/xapian/queryparser.h: Add documentation comments to NumberValueRangeProcessor. * tests/queryparsertest.cc: Update value_range2_queries to expect new serialised forms in its output. Update test_qp_value_range3 to test a wider range of values (including negative, fractional, and more-than-single-digit). Mon Jun 18 03:36:38 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix comment typo, and add brackets to make intended precedence clearer. Mon Jun 18 03:36:04 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Tweak wording of a couple of documentation comments. Mon Jun 18 02:00:31 BST 2007 Olly Betts <olly@survex.com> * docs/queryparser.html: Mention "AND NOT" as an alternative way to write "NOT". Mon Jun 18 01:43:00 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Adjust precedence of boolean operators to match those in maths and programming languages. "NOT" now binds as tightly as "AND" (previously "AND NOT" would bind like "AND", but just "NOT" would bind like "OR"!) Also "XOR" now binds more tightly than "OR", but less tightly than "AND" (previously it bound just like "OR"). * tests/queryparsertest.cc: Add testcases for the new behaviour. Mon Jun 18 01:16:18 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Better test coverage of handling of random double quotes in queries. Sun Jun 17 16:17:34 BST 2007 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Don't call operator++(0) from operator() as it does more work than we need. Sun Jun 17 14:14:16 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-inspect.cc: Fix so that Ctrl+D doesn't cause an infinite loop! Sat Jun 16 22:19:11 BST 2007 Olly Betts <olly@survex.com> * bin/Makefile.mk,bin/xapian-inspect.cc,xapian-core.spec.in: New utility 'xapian-inspect' allowing interactive inspection of key/tag pairs in a flint Btree. Useful for development and debugging, and an approximate replacement for quartzdump. Sat Jun 16 17:03:57 BST 2007 Olly Betts <olly@survex.com> * docs/doxygen_api_conf.in,docs/doxygen_full_conf.in: Update for doxygen 1.5.2 (using "doxygen -u"). Sat Jun 16 10:17:32 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am: The "check-none" target should run all the tests which don't need a database, not just those in apitest: add internaltest, stemtest, queryparsertest and termgentest to the list of tests it runs. Sat Jun 16 09:36:10 BST 2007 Richard Boulton <richard@lemurconsulting.com> * Makefile.am,bin/Makefile.mk,examples/Makefile.mk: Work around an apparent bug in automake which causes the entries in .libs subdirectories generated for targets of bin_PROGRAMS not to be removed on make clean. (This was causing make distcheck to fail.) Sat Jun 16 04:42:16 BST 2007 Olly Betts <olly@survex.com> * HACKING,Makefile.am,tests/Makefile.am,tests/apitest.cc, tests/harness/backendmanager.cc,tests/harness/backendmanager.h: In the testsuite, rename the "void" pseudo-backend to "none" to more clearly reflect what it is. And rename the "remote" backend to "remoteprog", to better differentiate it from "remotetcp". * HACKING: Fix error - "apitest -b=flint" doesn't work - it must be "apitest -bflint". Sat Jun 16 04:40:28 BST 2007 Olly Betts <olly@survex.com> * HACKING: Now using doxygen 1.5.2. Sat Jun 16 02:08:20 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Fix to build with older zlib. Not hugely important as zlib security fixes mean any well maintained box will be running a recent zlib, but the fix is trivial. Fri Jun 15 19:39:03 BST 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Document the new check-void, check-inmemory, check-remoteprog and check-remotetcp targets. Fri Jun 15 19:34:13 BST 2007 Richard Boulton <richard@lemurconsulting.com> * Makefile.am,tests/Makefile.am: Add check-void, check-inmemory, check-remoteprog and check-remotetcp targets, to allow the tests for any of the backends to be performed in isolation. Fri Jun 15 19:29:19 BST 2007 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,xapian-core.spec.in,xapian.spec.in: Rename xapian.spec to xapian-core.spec to match tarball name. Append the user name to BuildRoot. Fri Jun 15 14:51:13 BST 2007 Richard Boulton <richard@lemurconsulting.com> * net/progclient.cc: Open the file handle pointing to /dev/null as O_WRONLY instead of O_RDONLY: doesn't affect whether the test passes (since nothing gets written to stderr in theory anyway), but makes more sense. Fri Jun 15 13:21:56 BST 2007 Richard Boulton <richard@lemurconsulting.com> * net/progclient.cc: Better fix - rather than leaving stderr open, we close it, and then reopen it pointing to /dev/null. Fri Jun 15 12:55:07 BST 2007 Richard Boulton <richard@lemurconsulting.com> * net/progclient.cc: Don't close stderr in the child before we call execve() - with this change, apitest passes under valgrind with the remote database backend: before it, the child seemed to be exiting without writing anything to the socket, causing the test to fail with a NetworkError (due to an unexpected EOF). Thu Jun 14 23:47:52 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Replace duplicated code with calls to existing method set_revision_number(new_revision). Thu Jun 14 18:51:03 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Stop including <sys/utsname.h> which we don't use here. Thu Jun 14 18:22:24 BST 2007 Olly Betts <olly@survex.com> * tests/Makefile.am: Fix typo from copy&paste. Thu Jun 14 16:46:45 BST 2007 Olly Betts <olly@survex.com> * docs/admin_notes.rst: Update for the "lazy table" changes. Correct the description of xapian-compact --multipass (only the postlist tables are merged in multiple passes), correct description of how to use xapian-check on a single table (no trailing '.' should be specified). Change the "how to upgrade 0.9.x flint databases" to talk about 1.0.y rather than 1.0.0. Thu Jun 14 16:25:55 BST 2007 Olly Betts <olly@survex.com> * backends/flint/: The value and position tables are now only created if there is anything to add to them. So if you never use document values, there's no value_DB, value.baseA, or value.baseB. This means the table doesn't need to be opened for searching (saving a file handle) and when flushing changes, we don't need to update baseA/baseB just to keep the revisions in step. The flint database version has been increased, but the new code will happily open and read/update flint databases from Xapian 1.0.0 and 1.0.1. * bin/xapian-check.cc,bin/xapian-compact.cc: Update to handle databases without value and/or position tables. * tests/testdata/flint-1.0.1/: Empty example of a flint 1.0.1 format database. * tests/Makefile.am,tests/api_db.cc: Add check that a flint 1.0.1 format database can be opened. Wed Jun 13 22:40:15 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/queryparser.h: Document set_stemmer() and set_stemming_strategy() better - in particular, make it clear that the default value for the stemming strategy is STEM_NONE. Wed Jun 13 17:16:52 BST 2007 Olly Betts <olly@survex.com> * HACKING: Wrap overlong line. Wed Jun 13 16:47:07 BST 2007 Olly Betts <olly@survex.com> * msvc/version.h.in: Remote backend is now supported in the MSVC build. Wed Jun 13 04:18:51 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc,backends/flint/flint_table.h: Eliminate dont_close_handle as it's always false. Wed Jun 13 02:53:28 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_btreebase.cc,backends/flint/flint_btreebase.h: Remove unused FlintTable_base constructor. Tue Jun 12 18:27:54 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Add fix to allow compilation on OpenBSD, which uses off_t instead of uLong for total_out in zlib.h. Remove Assert() which is followed by a better check of the same condition. * AUTHORS: Add Julien Touche for help with this fix. Tue Jun 12 13:34:59 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/unixcmds.cc: On unix builds, set access permissions on copied directories to allow writes. This fixes make distcheck, which was failing on tests because the copied directory was set to read-only, causing the later "rm -rf" command to fail. Mon Jun 11 20:10:26 BST 2007 Olly Betts <olly@survex.com> * docs/overview.html: Fix typo. Mon Jun 11 20:08:42 BST 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Ship and install internals.html. Mon Jun 11 17:46:04 BST 2007 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/doxygen_full_conf.in: Disable generation of PDF version of doxygen-collated documentation of internals. It keeps overflowing TeX limits and it seems likely that most people would prefer the HTML version of this anyway. Mon Jun 11 17:45:12 BST 2007 Olly Betts <olly@survex.com> * tests/internaltest.cc: Remove duplicated code. Mon Jun 11 03:55:53 BST 2007 Olly Betts <olly@survex.com> * NEWS: Probably the final update for 1.0.1. Mon Jun 11 03:47:37 BST 2007 Olly Betts <olly@survex.com> * examples/simpleindex.cc: Tweak the logic to be clearer. Sun Jun 10 21:59:21 BST 2007 Olly Betts <olly@survex.com> * HACKING,configure.ac: Drop automake requirement to 1.8.3 to allow RPM spec file to work on SLES 9. Sun Jun 10 21:48:51 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Bump version to 1.0.1 and LIBRARY_VERSION_INFO to 15:0:0. Sun Jun 10 21:46:57 BST 2007 Olly Betts <olly@survex.com> * xapian.spec.in: Add "# norootforbuild" comment which SuSE's build scripts look for. Rename "Source0:" to "Source:" as there's only one tarball now. Add gcc-c++ and zlib-devel to "Build-Requires:". Sun Jun 10 21:43:45 BST 2007 Olly Betts <olly@survex.com> * PLATFORMS: More updates. Sun Jun 10 18:01:19 BST 2007 Olly Betts <olly@survex.com> * api/error.cc,generate-exceptions.in: Make Error::error_string member std::string rather than char * to avoid problems with double free() with copied Error objects. Also, Error::get_description() now converts my_errno to error_string if it hasn't been already. * tests/internaltest.cc: Add regression tests for both bugs. Sun Jun 10 13:33:21 BST 2007 Olly Betts <olly@survex.com> * tests/api_db.cc: New testcase in allterms6. Sun Jun 10 13:20:37 BST 2007 Olly Betts <olly@survex.com> * xapian-config.in: Add special case check for host_os matching linux* or k*bsd-gnu since vanilla libtool doesn't identify them as needing link_all_deplibs_CXX=no. Sun Jun 10 03:32:54 BST 2007 Olly Betts <olly@survex.com> * PLATFORMS: Updated from tinderbox. Sun Jun 10 02:14:58 BST 2007 Olly Betts <olly@survex.com> * NEWS: Updated. Sat Jun 09 17:28:32 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Add the unstem entry for boolean prefixed terms in Term::make_term() as we do for other terms. Sat Jun 09 17:27:55 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Replace reference to Xapian "1.0" with "1.0.0". Sat Jun 09 17:12:44 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,queryparser/queryparser_internal.h: Factor out the code which actually parses a term into a separate method. Fri Jun 08 23:31:23 BST 2007 Olly Betts <olly@survex.com> * tests/Makefile.am: Distribute tests/testdata/flint-0.9.9/. Fri Jun 08 19:20:36 BST 2007 Olly Betts <olly@survex.com> * NEWS: Updated. Fri Jun 08 18:55:43 BST 2007 Olly Betts <olly@survex.com> * examples/simplesearch.cc: Report "Matches 1-<N>:". Fri Jun 08 18:46:02 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/omenquire.cc: Fix issue brought to my attention by Enrico Zini: if an uninitialised database is supplied to Enquire, performing a search will cause a segfualt. It isn't completely clear whether using an uninitialised database should cause an error, or whether the search should return an empty result list, but for ease of implementation, and to avoid application errors going unnoticed, I've made Enquire's constructor report an InvalidArgumentError if the supplied database is uninitialised. * include/xapian/enquire.h: Document new exception case, and workaround if you want a search in this situation to return an empty result set, instead of getting an error. * tests/api_nodb.cc: Regression test for this fix. * AUTHORS: Mention Enrico Zini. Fri Jun 08 18:33:19 BST 2007 Olly Betts <olly@survex.com> * NEWS: Updated. Fri Jun 08 17:09:22 BST 2007 Olly Betts <olly@survex.com> * common/Makefile.mk: Distribute common/stringutils.h. Thu Jun 07 20:44:36 BST 2007 Olly Betts <olly@survex.com> * tests/harness/unixcmds.cc: It appears we need to create the target directory for xcopy to behave as we want. Thu Jun 07 19:34:18 BST 2007 Olly Betts <olly@survex.com> * tests/harness/unixcmds.cc: Fix typo to unbreak testsuite on Windows NT derivatives. Thu Jun 07 19:02:36 BST 2007 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_alltermslist.cc: Another tweak - don't call map::lower_bound() in the ctor at all. Also, prevent skip_to() from moving backwards, since that behaviour is user-visible. Thu Jun 07 18:36:49 BST 2007 Olly Betts <olly@survex.com> * common/stringutils.h: New header, providing inline functions begins_with() and ends_with() to check if a std::string has a given prefix or suffix. * api/valuerangeproc.cc,backends/flint/flint_alltermslist.cc, backends/inmemory/inmemory_alltermslist.cc, backends/quartz/quartz_alltermslist.cc, backends/quartz/quartz_metafile.cc,tests/harness/testsuite.cc: Make use of begins_with() and ends_with(). * tests/harness/testsuite.cc: Tidy up guessing of srcdir. In particular, fix comments to reflect that libtool won't create a wrapper script for the test programs on most platforms (because we now use --no-install when linking). * backends/inmemory/inmemory_alltermslist.cc: In the ctor, use std::map::lower_bound() to find the first term matching the given prefix, rather than iterating through from the start of the map. Also, don't skip deleted terms here, but instead do that in next() - this avoids duplicating code, and the first operation might be skip_to() in which case we can avoid wasting effort. Thu Jun 07 17:16:12 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Eliminate the QpQuery class and simplify various bits of the code. Thu Jun 07 14:32:22 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Enhance qp_value_customrange1. Thu Jun 07 14:31:40 BST 2007 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/index.html,docs/valueranges.rst: Add document describing how to use Xapian::ValueRangeProcessor. Thu Jun 07 14:16:01 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add test of custom ValueRangeProcessor subclass (qp_value_customrange1). Thu Jun 07 10:01:46 BST 2007 Olly Betts <olly@survex.com> * common/serialise.h,tests/harness/testsuite.h: Use XAPIAN_NORETURN() to mark functions which never return. Thu Jun 07 01:28:00 BST 2007 Olly Betts <olly@survex.com> * NEWS: Update. Thu Jun 07 01:02:22 BST 2007 Olly Betts <olly@survex.com> * examples/simpleexpand.cc: Fix off-by-one error when creating the fake RSet (picked up to 4 documents, not up to 5). Wed Jun 06 17:17:05 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Add documentation comments for ValueRangeProcessor and subclasses (mostly fixes bug#155, but I'd like to add an overview document too). Wed Jun 06 16:04:08 BST 2007 Olly Betts <olly@survex.com> * tests/testdata/flint-0.9.9/: Minimal example of a Flint database from Xapian 0.9.9. * tests/api_db.cc: Use a real 0.9.9 flint database instead of faking it by fudging with the "iamflint" file in a current one. flintdatabaseformaterror1 now also checks for opening with the Database ctor as well as with Flint::open(). The writing test from flintdatabaseformaterror1 is now flintdatabaseformaterror2 which additionally tests opening with the WritableDatabase ctor as well as with Flint::open(). The old flintdatabaseformaterror2 is now flintdatabaseformaterror3. Wed Jun 06 15:44:08 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Don't check for pre-0.6 databases here (since they'll be quartz format anyway - the check is just an overenthusiastic cut-and-paste from quartz). Wed Jun 06 12:35:16 BST 2007 Richard Boulton <richard@lemurconsulting.com> * matcher/valuerangepostlist.cc: Avoid calling db->get_lastdocid() repeatedly in ValueRangePostList::next() - was taking 25% of the time for a particular test case. Wed Jun 06 12:17:51 BST 2007 Olly Betts <olly@survex.com> * tests/harness/unixcmds.cc,tests/harness/unixcmds.h: Add "cp_R()" which does much the same as "cp -R" on Unix. Wed Jun 06 12:12:50 BST 2007 Olly Betts <olly@survex.com> * include/xapian/Makefile: New stub makefile for include/xapian. * bin/Makefile.mk,examples/Makefile.mk,include/Makefile.mk: Distribute missing stub Makefile and dir_contents files. Tue Jun 05 16:44:51 BST 2007 Olly Betts <olly@survex.com> * tests/harness/index_utils.cc: Back out linger line from r7758. Fix undefined behaviour in cases where a paragraph has <= 2 characters. Mon Jun 04 15:31:08 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/backendmanager.cc,tests/harness/index_utils.cc: Revert the recent changes to the indexing code, and also the change from revision r7758 which was intended to fix a problem with eof handling on windows (which is what broke the indexing of etext.txt). Then, change backendmanager.cc to test "!from.eof()" instead of "from" to check for eof - hopefully this will work on windows - if not, we can investigate further. Mon Jun 04 13:13:29 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/backendmanager.cc: Undo previous patch - only the patch to index_utils.cc was necessary; I mistakenly committed both files I'd been modifying. Mon Jun 04 09:06:24 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_db.cc: Add a check to consistency1 that the full mset size is as big as it should be, to make sure that we notice in future if the indexing goes wrong again. Mon Jun 04 08:59:22 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/backendmanager.cc,tests/harness/index_utils.cc: Fix the indexing of the testdata, so that the "etext" example is properly indexed (previously, only the first line was being indexed, due to faulty end-of-file handling). This only affects the "consistency1" test, which now runs a lot slower (but still passes, fortunately). Fri Jun 01 23:47:42 BST 2007 Olly Betts <olly@survex.com> * examples/simpleexpand.cc: Rewrite based on new simplesearch.cc. Fri Jun 01 00:50:00 BST 2007 Olly Betts <olly@survex.com> * examples/simpleindex.cc: Replacement "simpleindex" example which uses the TermGenerator class, which makes for a much smaller and simpler example. It's also much more typical of what most users will want to do. Thu May 31 23:01:47 BST 2007 Olly Betts <olly@survex.com> * examples/simplesearch.cc: Replacement "simplesearch" example which uses the QueryParser. That makes for a simpler example, and it's much more typical of what most users will want to do. Thu May 31 19:31:40 BST 2007 Olly Betts <olly@survex.com> * NEWS: Update. Thu May 31 19:18:41 BST 2007 Olly Betts <olly@survex.com> * HACKING,configure.ac: Relax automake requirement to 1.9.2 to allow RPM building on RHEL 4. * HACKING: automake 1.10 seems to work fine with Xapian. Thu May 31 03:04:42 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/quartz/quartz_database.cc: Check if the docid counter wraps and if it does throw DatabaseError (fixes bug#152). * tests/api_wrdb.cc: Add regression test (nomoredocids1). Thu May 31 02:01:31 BST 2007 Olly Betts <olly@survex.com> * backends/dbfactory_remote.cc,backends/remote/remote-database.cc, common/progclient.h,common/remote-database.h,common/tcpclient.h, net/progclient.cc: Stop a non-writable remote database from calling dtor_called() when it is destroyed, since that causes MSG_FLUSH to be sent to the server (fixes bug#149). Wed May 30 19:47:43 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/query.h: New accessor method Xapian::Query::Internal::get_parameter(). * queryparser/queryparser.lemony: Multiple boolean prefixed terms with the same term prefix are now combined with OR before such groups are combined with AND. Similarly for multiple value ranges on the same value (fixes bug#157). * include/xapian/queryparser.h: Document new handling of multiple boolean prefixed terms. * tests/queryparsertest.cc: Testcases for new handling of multiple boolean prefixed terms and value ranges. Wed May 30 14:45:19 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Update comment about QueryParserError to point out that you'll need to compile conditionally to work with 0.9.x Wed May 30 14:33:29 BST 2007 Olly Betts <olly@survex.com> * NEWS: Update. Put all the subheadings in the standard order used for previous releases. Wed May 30 12:58:01 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/database.h,api/omdatabase.cc,backends/flint/, backends/inmemory/,backends/quartz/, backends/remote/remote-database.cc, common/database.h,common/remote-database.h,common/remoteprotocol.h, net/remoteserver.cc: Add a prefixed form of allterms_begin() and allterms_end(), which allows efficient iteration over only those terms which begin with the given prefix. Requires bumping the remote protocol minor version number, but doesn't change any database formats. * queryparser/queryparser.lemony: Update the wildcard and partial query routines to use the prefixed form of allterms_begin(). This fixes the performance problems noted in bug #153. * tests/api_db.cc: Add test_allterms6() to test the prefixed form of allterms iterators. Wed May 30 12:23:25 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/remote/remote-database.cc,common/remoteprotocol.h, docs/remote_protocol.html,net/remoteserver.cc: Apply patch from Olly (from bug #153) which implements minor version numbers for the remove protocol. This will allow us to add new features to the protocol in future without forcing all clients to upgrade to exactly the same version as the servers. Bump the major version to 30, since this is an incompatible change itself, though. Wed May 30 12:12:14 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/query.h: Expand documentation of value range query constructor. Wed May 30 10:28:39 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Improve entry for QueryParserError. Wed May 30 01:30:35 BST 2007 Olly Betts <olly@survex.com> * backends/flint/,bin/xapian-compact.cc: Don't uncompress and recompress tags when compacting a database. This speeds up xapian-compact rather a lot (by more than 50% in my quick test). Tue May 29 20:58:13 BST 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/Makefile.mk: Fix multiple target rule for generating the queryparser source files in parallel builds, as described in HACKING, and as done for the SWIG rules in xapian-bindings. * queryparser/: Add queryparser_internal.lock and queryparser_internal.stamp to SVN ignores. Tue May 29 18:30:25 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/queryparsertest.cc: Use TEST_STRINGS_EQUAL when comparing queries against their expected output, since this makes it much easier to see the differences. Tue May 29 18:13:26 BST 2007 Richard Boulton <richard@lemurconsulting.com> * NEWS: Bring up-to-date. Mon May 28 09:27:43 BST 2007 Olly Betts <olly@survex.com> * examples/copydatabase.cc: Fix to compile with GCC 2.95. Mon May 28 01:28:23 BST 2007 Olly Betts <olly@survex.com> * examples/copydatabase.cc: Rewrite to take advantage of being able to iterate over all the documents in a database. Sun May 27 12:20:03 BST 2007 Olly Betts <olly@survex.com> * include/xapian/termgenerator.h: Improve documentation comments for TermGenerator class. Sun May 27 12:15:02 BST 2007 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc,bin/,examples/,net/tcpserver.cc, tests/api_db.cc,tests/harness/testsuite.cc,tests/harness/testutils.h, tests/queryparsertest.cc,tests/termgentest.cc: Make use of Xapian::Error::get_description(), which improves error reporting in a number of places. * tests/harness/testsuite.cc: If we get an exception with a long message in verbose mode, just report the whole message rather than reporting the truncated message followed by the whole message. * tests/queryparsertest.cc: Handle Xapian::QueryParserError specially (as we previously handled all Xapian::Error exceptions). Other Xapian::Error exceptions now report the description not just the message. Sun May 27 11:33:11 BST 2007 Olly Betts <olly@survex.com> * api/error.cc,generate-exceptions.in: Add new API method Xapian::Error::get_description() to allow removal of essentially duplicated code in many places. Also useful for users for the same reason. Sun May 27 09:56:21 BST 2007 Olly Betts <olly@survex.com> * generate-exceptions.in: Add vim modeline to force perl syntax highlighting. Sat May 26 22:53:52 BST 2007 Richard Boulton <richard@lemurconsulting.com> * NEWS: Bring up-to-date. Sat May 26 22:49:20 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Fix typo in deprecation list for the bindings pointed out by Thomas Waldmann. Sat May 26 09:39:36 BST 2007 Olly Betts <olly@survex.com> * HACKING: Add note about reasons for avoiding std::pair<> with an STL class as one or both members. Thu May 24 17:28:20 BST 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/Makefile.mk: $(top_builddir)/queryparser is only required in include path for maintainer mode, so make it so for neatness. Tue May 22 11:04:04 BST 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/Makefile.mk: Add $(top_builddir)/queryparser to include path for VPATH builds, to fix a problem with depcomp (triggered when compiling with gcc-2.95). * NEWS: Keep up to date. Mon May 21 13:38:58 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/queryparsertest.cc: Reduce the range of the test, so that it passes again, to let the autobuilders work and unbreak HEAD. Mon May 21 10:26:01 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/queryparsertest.cc: Remove FIXME from test_qp_flag_partial1: I've checked that the results look correct. New test test_qp_value_range3, to test the NumberValueRangeProcessors - currently fails because "10" is considered to be between "1" and "2". * include/xapian/query.h: Fix typo in documentation comment. Mon May 21 10:09:34 BST 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.h: Put quotes around strings when reporting a failure in TEST_STRINGS_EQUAL(). Mon May 21 10:08:18 BST 2007 Olly Betts <olly@survex.com> * tests/internaltest.cc: New tests serialiseerror1 which checks that the handling of errno/error_string. * common/serialise.h: Make unserialise_error externally visible for new test. Mon May 21 08:40:06 BST 2007 Olly Betts <olly@survex.com> * HACKING: Update release checklist. Fri May 18 02:10:39 BST 2007 Olly Betts <olly@survex.com> * generate-exceptions.in: Fix grammatical errors in Error::get_errno() deprecation explanation. Thu May 17 23:51:17 BST 2007 Olly Betts <olly@survex.com> * api/error.cc: Need <stdlib.h> for free() to build with GCC 4.3 snapshot. Thu May 17 23:11:53 BST 2007 Olly Betts <olly@survex.com> * NEWS: Fix line wrapping (ok, isn't wasn't final!) Thu May 17 22:28:44 BST 2007 Olly Betts <olly@survex.com> * NEWS: Final update before release. Thu May 17 19:10:27 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Bump version to 1.0.0 and update LIBRARY_VERSION_INFO. Thu May 17 19:10:11 BST 2007 Olly Betts <olly@survex.com> * NEWS: Minor wording tweak. Thu May 17 19:06:34 BST 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Run rst2html with --strict --verbose options to reject .rst files with problems. * docs/deprecation.rst: Add more entries gleaned from the NEWS file. Thu May 17 18:30:01 BST 2007 Olly Betts <olly@survex.com> * NEWS: Update with fixes from Jenny's proof-reading and a few other tweaks. Thu May 17 18:27:53 BST 2007 Olly Betts <olly@survex.com> * AUTHORS: Update. Thu May 17 18:25:39 BST 2007 Olly Betts <olly@survex.com> * PLATFORMS: Updates from the tinderbox. Thu May 17 18:08:13 BST 2007 Olly Betts <olly@survex.com> * common/safesysstat.h: Need the 2 argument mkdir() wrapper for Mingw too. Define S_ISDIR() and S_ISREG() if they aren't already defined rather than conditional on _MSC_VER. Thu May 17 17:58:43 BST 2007 Olly Betts <olly@survex.com> * docs/admin_notes.rst: "flint-check" -> "xapian-check". Thu May 17 17:46:36 BST 2007 Olly Betts <olly@survex.com> * NEWS: Update for Xapian 1.0.0. Thu May 17 17:41:15 BST 2007 Olly Betts <olly@survex.com> * docs/: Update documentation for 1.0.0; add new document describing new term generation strategy as implemented by TermGenerator class; link in "lost" documents to the index page. Thu May 17 03:54:36 BST 2007 Olly Betts <olly@survex.com> * exception_data.pm: Fix to work with Perl 5.6. Thu May 17 01:20:16 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/multi/multi_alltermslist.cc: Check that iterators are not at_end() before calling skip_to() on them. Thu May 17 00:49:03 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_btreebase.cc,backends/flint/flint_table.cc, backends/quartz/btree.cc,common/Makefile.mk, common/msvc_posix_wrapper.cc,common/msvc_posix_wrapper.h: Use msvc_posix_wrapper.h functions for all __WIN32__ builds, not just for MSVC. Wed May 16 23:55:46 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc: Fix return type in debug macro in replace_document(). Wed May 16 23:45:20 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/apitest.cc: Add collfreq and allterms tests to the tests for the remote and remotetcp backends. They all pass. Wed May 16 15:11:00 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Need to try -zdll for mingw as well. Wed May 16 14:45:23 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add two more testcases. Wed May 16 14:32:46 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Fix double-free on invalid range query. Wed May 16 10:08:02 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/queryparsertest.cc: Add some more test cases to the value range processor. These currently fail with a SIGSEGV message. Wed May 16 02:22:17 BST 2007 Olly Betts <olly@survex.com> * queryparser/termgenerator_internal.cc: Fix handling of stemming of prefixed terms. * tests/termgentest.cc: Fix testcases. Wed May 16 01:54:25 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/termgentest.cc: Add a test for parsing of capitalised words with a prefix, which currently fails. Wed May 16 01:09:39 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/termgenerator.h,queryparser/termgenerator.cc: Add a method to get the current document from the term generator. Helps avoid users having to pass around a structure encapsulating the generator and the document assigned to it. Wed May 16 00:50:07 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Add <xapian/output.h>. Tue May 15 22:46:39 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/termgentest.cc: Use resize(0) instead of clear(), for g++2.95 Tue May 15 20:25:07 BST 2007 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: Add a comment noting why we test __STDC_SECURE_LIB__ even though its an undocumented macro. Tue May 15 20:14:51 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/termgentest.cc: Add several more tests - some converted from the query tests, some more initial tests, and some number tests. Tue May 15 18:31:21 BST 2007 Olly Betts <olly@survex.com> * queryparser/termgenerator_internal.cc: Revert previous change - we need to generate Zibm and Zpc from text `I.B.M. P.C.' or else a search for `ibm pc' won't match. Tue May 15 18:22:44 BST 2007 Olly Betts <olly@survex.com> * queryparser/termgenerator_internal.cc: Sort out stemming of initials. Tue May 15 18:04:17 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/termgentest.cc: Add some more tests, and rework the array of tests so that there's just a single string holding any options we want to set for the following tests. Tue May 15 16:49:17 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/Makefile.am,tests/termgentest.cc: Add new test file, for testing the term generator. Only has a single test so far. Tue May 15 16:20:20 BST 2007 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Make single parameter ctor explicit. Tue May 15 16:00:24 BST 2007 Olly Betts <olly@survex.com> * queryparser/termgenerator_internal.cc: Add special handling for acronyms/initialisms, so I.B.M -> term ibm, etc. Tue May 15 14:07:27 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/termgenerator.h,queryparser/termgenerator.cc: Add a set_termpos() method, too. Tue May 15 13:54:34 BST 2007 Olly Betts <olly@survex.com> * include/xapian/database.h: Document the issue of the monotonic docid counter wrapping around. Tue May 15 10:58:37 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/termgenerator.h,queryparser/termgenerator.cc: Add get_termpos() method to get the current value of the termpos. Makes it easier to integrate with different term generators. Tue May 15 03:44:50 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,queryparser/termgenerator_internal.cc: Include '.', ',', and a few other characters in terms if they are between two decimal digits. * queryparser/termgenerator_internal.cc: Don't stem terms which start with a number. * tests/queryparsertest.cc: Update results for real world queries to match new results (better in every case but one!) Tue May 15 01:24:08 BST 2007 Olly Betts <olly@survex.com> * HACKING: Fix a few typos. Add "feisty" to the list of debs to build. Tue May 15 00:51:35 BST 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Document that the bootstrap script should be run from srcdir, and that the generated configure can be run from a different builddir. Tidy up a few other lines. Mon May 14 23:52:33 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,queryparser/termgenerator_internal.cc: Expand check_infix to include some of characters Unicode does. * tests/queryparsertest.cc: Update the single test query which now parses differently. Mon May 14 23:42:53 BST 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.h: New macro TEST_STRINGS_EQUAL() which displays the strings on separate lines so the differences can be clearly seen. * tests/queryparsertest.cc: Use TEST_STRINGS_EQUAL(). Mon May 14 20:14:17 BST 2007 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Make the "uses string sort" warning in the documentation for the methods which enable sorting by value more prominent, and fix several miscellaneous typos. Mon May 14 19:28:21 BST 2007 Olly Betts <olly@survex.com> * queryparser/dir_contents: Update to mention TermGenerator. Mon May 14 19:12:12 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix handling of WILDCARD_TERM and PARTIAL_TERM for prefixed terms. * tests/queryparsertest.cc: Add test coverage for this. Mon May 14 18:02:22 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add more testcases for wildcarded queries. Mon May 14 17:39:30 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix handling of partial queries. Mon May 14 16:49:46 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: closesocket() seems to work equally as well as CloseHandle() but it's what we use elsewhere and Charlie thinks it's a better choice here too so use that instead. Rename the wrapper call from close_fd_or_handle() to close_fd_or_socket(). And mark close_fd_or_socket() as "inline" under UNIX, since it should always be inlined. Mon May 14 15:30:13 BST 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Add FIXME about moving wildquery1. Mon May 14 15:21:59 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-progsrv.cc: Improve comments. Mon May 14 11:29:14 BST 2007 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: Deal with invalid parameter exceptions on windows when closing file descriptors or handles. Mon May 14 05:04:16 BST 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Fix wildquery1 to work with the new term stemming scheme. Sun May 13 19:15:17 BST 2007 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Count control characters as whitespace (which is certainly what we want for TAB, CR, LF, and FF). Sun May 13 17:50:19 BST 2007 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Add CONNECTOR_PUNCTUATION to is_wordchar() - this category includes '_' and characters which fulfil similar roles. * queryparser/queryparser.lemony,queryparser/termgenerator_internal.cc: No longer need to special case '_'. * queryparser/queryparser.lemony: Test for Unicode "whitespace" rather than just ASCII. Sun May 13 04:43:37 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Normalise Unicode characters used for apostrophe to ASCII apostrophe. Sun May 13 03:56:17 BST 2007 Olly Betts <olly@survex.com> * include/xapian/unicode.h,unicode/tclUniData.cc: Update Unicode routines to use Unicode 5.0.0 and support characters outside the BMP (so we now support all Unicode characters in the latest version of the standard). * tests/api_unicode.cc: Add test unicode1 which performs some simple tests of the categorisation function, include one which tests for a character added in Unicode 5.0.0. Sat May 12 06:01:50 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Update comment. Sat May 12 04:23:01 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Improve description of Term class. Sat May 12 04:16:47 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Update to implement the new stemming scheme. * tests/queryparsertest.cc: Update testcases, and check that the new parses are all reasonable. Sat May 12 04:03:55 BST 2007 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Xapian::Unicode::get_category() now checks for characters outside the BMP, and assumes they're of category OTHER_LETTER. Remove BMP check from all functions which use Xapian::Unicode::get_category(). Sat May 12 04:02:52 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Fix typo in comment. Fri May 11 05:45:38 BST 2007 Olly Betts <olly@survex.com> * net/remoteserver.cc: Make "Server is read-only" be of type InvalidOperationError not NetworkError. This is arguably more appropriate anyway, but in particular it avoids a worrying looking (but harmless) warning when a read-only connection is closed, due dtor_called() calling flush() on a read-only database. Fri May 11 01:35:31 BST 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Disable more documentation rules if MAINTAINER_NO_DOCS is true. Thu May 10 23:01:13 BST 2007 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: Remove definition of unused macro XAPIAN_MSVC_INVALID_PARAMETER_HANDLER. Thu May 10 19:24:23 BST 2007 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Try quoting qualified identifiers in #pragma deprecated. Thu May 10 15:58:37 BST 2007 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Give full scope in '#pragma deprecated' to try to avoid false positives. Thu May 10 14:53:08 BST 2007 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Use '#pragma deprecated' to work around MSVC stupidity (you get a warning for deprecating a declaration). Thu May 10 13:42:49 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Some GCC 3.x versions produce bogus warnings, so only automatically enable -Werror on --enable-maintainer-mode under GCC if we're using GCC >= 4.0. Most developers will be using GCC 4.x soon if they aren't already anyway. * HACKING: Update documentation in line with the above. Mention that we also enable -Werror for Intel's C++ compiler. Thu May 10 05:06:22 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Detect __WIN32__ by running the preprocessor so we don't get confused by building mingw in a cygwin environment. Thu May 10 04:52:21 BST 2007 Olly Betts <olly@survex.com> * net/tcpserver.cc: Exit with EX_UNAVAILABLE if the mutex tells us xapian-tcpsrv is already running on this port. Thu May 10 04:44:10 BST 2007 Olly Betts <olly@survex.com> * common/Makefile.mk,common/safe.cc,common/safeunistd.h: Pulling in safewindows.h from safeunistd.h causes clashes with NEAR in the QueryParser, so use a helper function to call Sleep and just call that from the header. Thu May 10 04:36:19 BST 2007 Olly Betts <olly@survex.com> * net/tcpserver.cc: Fix compilation error. Thu May 10 04:26:31 BST 2007 Olly Betts <olly@survex.com> * common/tcpserver.h,net/tcpserver.cc: Just pass the mutex to get_listening_socket() by reference to avoid MSVC warnings. Thu May 10 04:19:13 BST 2007 Olly Betts <olly@survex.com> * common/safeunistd.h: Use Sleep() instead of _sleep() as the latter gives deprecation warnings. Sadly that means that safeunistd.h now has to pull in safewindows.h. Thu May 10 04:07:53 BST 2007 Olly Betts <olly@survex.com> * common/tcpserver.h,net/tcpserver.cc: Keep track of the mutex so we can release it right after we close the listening socket. Throw an error if we fail to get the mutex to avoid a possible race condition (we might end up with the socket listening but without the mutex locked). Always call closesocket() on the listening socket under __WIN32__ to ensure resources are released. Thu May 10 02:48:31 BST 2007 Olly Betts <olly@survex.com> * net/tcpserver.cc: Use a mutex under Cygwin and WIN32 to allow use to use Microsoft's broken SO_REUSEADDR on our listening socket, but avoid being able to run two copies of xapian-tcpsrv on the same port at once (tweaked patch from MarkH). * net/tcpserver.cc: Don't try to set SO_EXCLUSIVEADDRUSE if a previous setsockopt() call failed as we might stomp on the error code. Thu May 10 02:15:20 BST 2007 Olly Betts <olly@survex.com> * Makefile.am,bin/Makefile.mk,configure.ac,docs/Makefile.am, examples/Makefile.mk: Add new Automake conditional "MAINTAINER_NO_DOCS" to allow use to disable references to built documentation in maintainer builds if --disable-documentation is specified. This allows Mark's buildbot to work without installing so many tools, some of which aren't packaged for Cygwin. Thu May 10 02:03:44 BST 2007 Olly Betts <olly@survex.com> * Makefile.am: Build rules for running generate-exceptions should be controlled by MAINTAINER_MODE not DOCUMENTATION_RULES, part II. Thu May 10 01:49:42 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Add Error::get_type(). Thu May 10 01:25:37 BST 2007 Olly Betts <olly@survex.com> * api/error.cc,generate-exceptions.in: Add new constructors which allow error_string to be set directly, for use by the Xapian::Error unserialising code. * generate-exceptions.in: Xapian::Error::get_type() now returns const char * not std::string. * backends/quartz/quartz_database.cc: Fix code to handle changed return type of Xapian::Error::get_type(). * net/serialise.cc: Include error_string in the serialisation. * common/remoteprotocol.h,docs/remote_protocol.html: Bump protocol version to 29 because of change in Xapian::Error serialisation. Wed May 09 14:42:33 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Renaming match_nothing_ member of QpQuery to match_nothing. Comment out unused QpQuery::get_description(). Add FIXME comment about apparently redundant logic (now's not really the time to be fiddling with such things). Wed May 09 14:27:02 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Update to match new TermGenerator class. '_' is now a term character, not a phrase generator. We only keep a suffix which has 3 or fewer characters, and we no longer count '-' as a suffix character. * tests/queryparsertest.cc: Update test cases to match new behaviour. In almost every case the new rules give a better result. Wed May 09 14:15:38 BST 2007 Olly Betts <olly@survex.com> * languages/compiler/space.c: Add sanity check to b_to_s(), highlighted by MSVC warning. Wed May 09 14:06:24 BST 2007 Olly Betts <olly@survex.com> * tests/btreetest.cc: Return filesize as off_t to avoid MSVC warning about truncating from 64 to 32 bits (the filesize will never be more than a few K here, so this is essentially a cosmetic issue). Wed May 09 04:04:03 BST 2007 Olly Betts <olly@survex.com> * common/tcpserver.h,net/tcpserver.cc: Always starting the server in a new thread in "one-shot" mode is trickier than it looked, so revert that change for now. Always call closesocket() on our sockets under __WIN32__. Wed May 09 01:53:43 BST 2007 Olly Betts <olly@survex.com> * common/tcpserver.h,net/tcpserver.cc: Make UNIX signal handling functions and WIN32 CrtlHandler static. TcpServer::run() is now generic, calling an OS-specific TcpServer::run_once(). Now we always run the server in a new thread in "one-shot" mode, which gives better code coverage in apitest as it runs more like a normal xapian-tcpsrv would. CtrlHandler now lets the OS handle the event if it doesn't have a socket to close, or if closesocket() fails. Wed May 09 01:42:33 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: DWORD is unsigned, so received can't be negative - tidy up the ReadFile() loop taking this into account. Wed May 09 01:28:48 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: ReadFile() doesn't set errno, so remove test for EINTR which is just a hangover from the UNIX code. Tue May 08 23:50:33 BST 2007 Olly Betts <olly@survex.com> * queryparser/termgenerator_internal.cc: Treat '_' as a word character which seems to be the correct thing to do in general. Generate unstemmed terms with positional information for all words without a prefix, and stemmed terms without positional information for non-stopwords with a 'Z' prefix. This means we can implement exact phrase search, and prefixing stemmed forms (rather than unstemmed as Omega < 1.0 does) means fewer terms with the prefix. Only stemming non-stopwords cuts a significant amount of the database size in my tests. Tue May 08 18:22:30 BST 2007 Olly Betts <olly@survex.com> * net/tcpserver.cc: Fix remaining use of handle_one_request in __WIN32__-specific code. Tue May 08 18:10:12 BST 2007 Olly Betts <olly@survex.com> * api/error.cc: Really fix it... Tue May 08 17:38:04 BST 2007 Olly Betts <olly@survex.com> * api/error.cc: Fix another paste-o! Tue May 08 17:28:29 BST 2007 Olly Betts <olly@survex.com> * api/error.cc: Remove bogus statement in __WIN32__ case left over from cut-and-paste. Tue May 08 17:10:53 BST 2007 Olly Betts <olly@survex.com> * common/tcpserver.h,net/tcpserver.cc: Rename TcpServer::handle_one_request() to TcpServer::handle_one_connection() (to make it clearer that it handles a connection rather than a single MSG/REPLY exchange), and use it in UNIX builds too to avoid duplicating its code. Catch NetworkTimeoutError here and report it if we're in verbose mode. * net/remoteserver.cc: If we catch NetworkTimeoutError then try to propagate it only if we can send it right away, then rethrow it to close the connection and let the caller log it. If we catch any other NetworkError then just rethrow it. Tue May 08 16:52:50 BST 2007 Olly Betts <olly@survex.com> * common/noreturn.h: Add missing '#endif'. Tue May 08 16:19:05 BST 2007 Olly Betts <olly@survex.com> * common/Makefile.mk,common/noreturn.h: New header which defines XAPIAN_NORETURN() to allow functions which don't return to be marked as such so the compiler can take this into account when generating code. * api/error.cc,tests/harness/backendmanager.cc: Remove any trailing "\r\n" from the output of FormatMessage(). Tue May 08 16:16:55 BST 2007 Olly Betts <olly@survex.com> * include/xapian/deprecated.h: Use `__attribute__((__deprecated__))' instead of `__attribute__((deprecated))' so we're even robust against some clown doing `#define deprecated foo'. Tue May 08 16:15:30 BST 2007 Olly Betts <olly@survex.com> * bin/quartzcheck.cc,tests/harness/backendmanager.cc: In these files we have to disable XAPIAN_DEPRECATED() since we need to be able to use Quartz::open() - move this to the start of the #include block to avoid problems with any intermediate header implicitly including <xapian.h>. Tue May 08 14:51:37 BST 2007 Olly Betts <olly@survex.com> * HACKING: Move section on "Configure Options" to a more logical spot just before "Makefile Portability". Add section on how to use XAPIAN_DEPRECATED() noting the issue with GCC 3.3.5 not allowing it to be used on an inline method definition. * include/xapian/deprecated.h: Add pointers to HACKING and docs/deprecated.rst. Tue May 08 13:57:17 BST 2007 Olly Betts <olly@survex.com> * generate-exceptions.in: We can't use XAPIAN_DEPRECATED() on a function definition with GCC 3.3.5, so use it on a declaration and then define inline but out of the class. Tue May 08 13:45:48 BST 2007 Olly Betts <olly@survex.com> * net/tcpserver.cc: We must call closesocket() (instead of just close()) under __WIN32__ or else the socket remains in the CLOSE_WAIT state. Tue May 08 12:17:19 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Add Error::get_errno() to the deprecation list for the bindings, too. Tue May 08 10:09:27 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Change the xapian_version functions to be marked for removal in 1.1.0, and also note them in the bindings section. Tue May 08 04:27:37 BST 2007 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remote-database.h: The timeout parameter to RemoteDatabase wasn't being used, instead the client would wait indefinitely for the server to respond. Tue May 08 04:16:45 BST 2007 Olly Betts <olly@survex.com> * generate-exceptions.in: Initialise error_string in Xapian::Error ctor. Tue May 08 03:48:47 BST 2007 Olly Betts <olly@survex.com> * api/error.cc: Fix compilation when HAVE_HSTRERROR isn't defined. Tue May 08 02:35:04 BST 2007 Olly Betts <olly@survex.com> * api/error.cc,bin/xapian-tcpsrv.cc,examples/delve.cc, generate-exceptions.in,tests/harness/testsuite.cc: Error subclasses now store the error_string in the class, converting it from the error code in my_errno lazily. Also, get_error_string() now returns const char *. Tue May 08 01:34:56 BST 2007 Olly Betts <olly@survex.com> * Makefile.am: Build rules for running generate-exceptions should be controlled by MAINTAINER_MODE not DOCUMENTATION_RULES. Tue May 08 01:23:22 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-tcpsrv.cc,docs/deprecation.rst,generate-exceptions.in, net/tcpserver.cc,tests/harness/testsuite.cc: Deprecate Error::get_errno() in favour of Error::get_error_string(). We can't pass errno values from remote server to client as they aren't portable and the server and client may be running on entirely different platforms! * generate-exceptions.in: Mark as generated with @configure_input@. Tue May 08 01:20:43 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_lock.h: Make FlintLock's dtor release the lock (if held) which avoids the need for any special handling in FlintDatabase. Mon May 07 21:15:33 BST 2007 Olly Betts <olly@survex.com> * tests/harness/unixcmds.cc: Use O_BINARY when opening a file to "touch" it. Mon May 07 17:28:43 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_database.cc: Add try{}catch(...){} block to ensure that lock is released if an exception occurs in the FlintDatabase() constructor. * tests/api_db.cc: Add regression test for releasing lock on error in constructor. Mon May 07 11:10:33 BST 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Update section on multiple output rules to detail the workaround described in the automake manual. Mon May 07 10:03:50 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_database.cc,backends/flint/flint_database.h, backends/flint/flint_version.cc,backends/quartz/,exception_data.pm, include/xapian/database.h,tests/api_db.cc: Throw DatabaseVersionError instead of DatabaseOpeningError when database version is unsupported. This is a new error class, which is a subclass of DatabaseOpeningError, so most existing scripts shouldn't need changing, but makes it easy for application writers to determine whether a database needs upgrading or whether the problem is something else. Should fix #144. Mon May 07 09:22:01 BST 2007 Olly Betts <olly@survex.com> * net/progclient.cc: Under __WIN32__, use byte mode on pipes because message mode fails for message > 256 bytes, and use "overlapped" mode so we don't block waiting for a reply so that timeouts work. (Patch from Mark Hammond). Mon May 07 03:03:10 BST 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc,tests/apitest.cc,tests/harness/: Now TEST_EQUAL_DOUBLE() should work everywhere. Move the helper function out of the header. Use TEST_EQUAL_DOUBLE() instead of doubles_are_equal_enough() and weights_are_equal_enough() which do much the same job. Mon May 07 01:13:29 BST 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.h: Report extra info to try to work out why MSVC build is failing expandweights1. Mon May 07 00:26:59 BST 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.h: Another tweak to TEST_EQUAL_DOUBLE(). Sun May 06 22:30:26 BST 2007 Olly Betts <olly@survex.com> * docs/quickstart.html: Remove reference to DA databases since the muscat36 backend has been removed. Update to recommend GCC 3.3 or later, as we do elsewhere. Change references to "quartz" to say "flint", or remove them. Sun May 06 16:02:28 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Set version to mythical 0.9.99. Sun May 06 15:06:01 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_db.cc: Add test_flintdatabaseformaterror1() which checks what happens when an old-format flint database is opened. It generates the old-format database by making a fake "iamflint" file containing an old version number. Sun May 06 04:00:39 BST 2007 Olly Betts <olly@survex.com> * expand/ortermlist.cc: Remove incorrect assertion check in OrTermList::next(), replacing it with anexplanatory comment. Sun May 06 02:42:30 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Fix typo. Sun May 06 02:39:23 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Aha! Richard's patch failed to compile because stdlib.h was in "#ifndef __WIN32__" not "#ifdef __WIN32__". Sun May 06 02:27:18 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Protect _set_invalid_parameter_handler, etc by _MSC_VER >= 1400 (MSVC 2005) and __STDC_SECURE_LIB__ (undocumented but cribbed from the Python sources). Refactor into a handy class (MSVCIgnoreInvalidParameter) which just needs to be instantiated in the scope where you want to allow invalid parameters to be passed. Sun May 06 01:42:06 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Rework use of _invalid_parameter_handler to only apply when _MSC_VER is defined, to fix the mingw build. Sat May 05 22:30:01 BST 2007 Richard Boulton <richard@lemurconsulting.com> * net/remoteconnection.cc: Under __WIN32__, add an invalid parameter handler which ignores the error and use it for the calls to _get_osfhandle() which are sometimes expected to fail. Requires stdlib.h according to msdn, so also #include that. Sat May 05 20:43:29 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-progsrv.cc,bin/xapian-tcpsrv.cc,common/remoteconnection.h, common/remoteserver.h,common/tcpserver.h,net/remoteserver.cc, net/tcpserver.cc: Open a fresh copy of the database(s) on each connection to a xapian-tcpsrv rather than relying on being able to share a database across fork() or between threads (which we don't promise will work). Largely based on patch from Mark Hammond. Fixes some of bug#141 and all of bug#142. Sat May 05 02:19:39 BST 2007 Olly Betts <olly@survex.com> * api/omdatabase.cc: Use RETURN() not return in methods with debug logging. Sat May 05 02:19:11 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Fix typo in comment. Fri May 04 23:59:54 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_version.cc: Remove the "u" suffix from FLINT_VERSION as it's confusing in the "wrong format version" error message and doesn't seem needed in the code. Fri May 04 23:56:32 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/: Revert patch from revision 8426 (except for the bit which removed the unused prototypes), to avoid the possibility that it introduces bugs into the code just before 1.0.0. The patch is attached to bug #143 for later reference. Fri May 04 23:50:39 BST 2007 Olly Betts <olly@survex.com> * examples/delve.cc: Report the error string if we catch Xapian::Error while trying to open a database. Fri May 04 22:50:54 BST 2007 Olly Betts <olly@survex.com> * docs/intro_ir.html: Fix a few typos; elaborate on a few points. Fri May 04 20:07:44 BST 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/termgenerator_internal.cc: Fix bug in termgenerator - arguments to add_posting were the wrong way round. Fri May 04 20:00:40 BST 2007 Olly Betts <olly@survex.com> * queryparser/termgenerator_internal.cc: Limit term length, handle infix and suffix characters, and convert apostrophes to ASCII representation. Fri May 04 19:47:43 BST 2007 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Fill in remaining missing documentation comments. Fri May 04 19:30:56 BST 2007 Olly Betts <olly@survex.com> * include/xapian/termgenerator.h: Add convenience version of index_text() and index_text_without_positions() which take a std::string instead of a Utf8Iterator. Fri May 04 17:11:11 BST 2007 Richard Boulton <richard@lemurconsulting.com> * unicode/utf8itor.cc: Fix segfault bug in Utf8Iterator::operator*(). If this was called when the iterator had reached end, invalid memory was accessed because p was compared to end, p will always be set to NULL when the iterator reaches the end. No regression test, but the smoketest tests I'm implementing in the bindings will exercise this code. Fri May 04 16:29:51 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/unicode.h: Add documentation comments to Utf8Iterator, so I don't have to work out how it works again. Fri May 04 14:55:32 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/xapian/unicode.h: Add assign() function which takes a string, for symmetry with the constructors. Also makes implementing the bindings easier. Fri May 04 13:04:52 BST 2007 Richard Boulton <richard@lemurconsulting.com> * include/Makefile.mk: Add termgenerator.h to distribution tarballs. Fri May 04 03:41:56 BST 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.h: Test against pow(10, -DBL_DIG) rather than DBL_EPSILON. DBL_EPSILON is a little bit smaller, but we use DBL_DIG for displaying the numbers and it looks stupid to say two numbers which are displayed identically should be nearly the same! Fri May 04 03:33:51 BST 2007 Olly Betts <olly@survex.com> * HACKING: Fix reference to --enable-maitainer-mode and documentation rebuilding. Fri May 04 03:32:57 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Fix now erroneous references to tools "required in maintainer mode" to "required to build documentation". Fri May 04 01:49:05 BST 2007 Olly Betts <olly@survex.com> * HACKING: Document --enable-documentation. Fri May 04 01:38:44 BST 2007 Olly Betts <olly@survex.com> * Makefile.am,bin/Makefile.mk,configure.ac,docs/Makefile.am, examples/Makefile.mk: Add new configure option "--enable-documentation" which enables the make rules to rebuild the documentation. By default this follows the setting of "--enable-maintainer-mode" (so giving much the same behaviour as currently), but the documentation rules can now be controlled independently. Also, if documentation isn't being built, the configure probes for tools required to rebuild it are no longer run which should shorten configure time a bit for end-users building releases. Thu May 03 22:07:39 BST 2007 Olly Betts <olly@survex.com> * common/remoteserver.h: Remove unused default values for active_timeout_ and idle_timeout_ parameters. Fix repeated typo in documentation comments ('millisrconds' -> 'milliseconds'). Thu May 03 15:05:24 BST 2007 Olly Betts <olly@survex.com> * include/xapian.h,include/xapian/termgenerator.h,queryparser/: Initial cut of TermGenerator class. Thu May 03 15:04:09 BST 2007 Olly Betts <olly@survex.com> * generate-exceptions.in: Hide the internal constructors of generated error classes in the doxygen-generated API docs. Thu May 03 04:37:51 BST 2007 Olly Betts <olly@survex.com> * expand/ortermlist.cc,expand/ortermlist.h: Fix compilation error when --enable-assertions is used. Thu May 03 03:57:42 BST 2007 Olly Betts <olly@survex.com> * common/safeunistd.h: Provide a POSIX-compatible sleep() function. * common/utils.h: Remove inferior POSIX-compatible sleep() function. Thu May 03 03:17:02 BST 2007 Olly Betts <olly@survex.com> * include/xapian/: Mark all @internal member variables and functions which aren't private with @private so doxygen doesn't document them. Thu May 03 02:33:09 BST 2007 Olly Betts <olly@survex.com> * include/xapian/: Mark DocIDWrapper, RefCntBase, RefCntPtr, TermNameWrapper, and TermPosWrapper as "@internal" so that they don't show up in the doxygen-generated API documentation. In particular, this makes the class collaboration diagrams much more useful. * include/xapian/errorhandler.h: Declare the class in namespace Xapian rather than with namespace Xapian:: since the latter seems to confuse doxygen. Thu May 03 02:31:33 BST 2007 Olly Betts <olly@survex.com> * include/xapian/unicode.h: Fix wrapping of comment. Thu May 03 00:05:23 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/: Remove unused (and unimplemented) declarations of begin_transaction_(), commit_transaction_() and cancel_transaction_() from FlintWritableDatabase class. Tidy up handling of metainfo (total_length and lastdocid) to avoid repeatedly parsing the special record, and make it easier to store additional metainfo in future. Also, remove code which ensures that there is always a special record - this was added in revision 5459 when the special record was stored in the record table, so that the number of documents in the database could always be calculated by subtracting 1 from the number of entries in the record table. Now that the special record is stored in the postlist table, this is no longer necessary. Wed May 02 17:56:39 BST 2007 Olly Betts <olly@survex.com> * backends/dbfactory_remote.cc,include/xapian/dbfactory.h: Remote::open_writable() now defaults to no (active) timeout. The connection timeouts for Remote::open() and Remote::open_writable() now default to 10 seconds rather than defaulting to the same as the active timeout (which defaults to 10 seconds). Wed May 02 16:00:22 BST 2007 Richard Boulton <richard@lemurconsulting.com> * common/safeunistd.h,tests/harness/backendmanager.cc: Fix compilation on windows - safeunistd.h was undefining ssize_t and thus overriding the definition in config.h, and include safewindows.h for backendmanager.cc Wed May 02 15:34:33 BST 2007 Olly Betts <olly@survex.com> * common/safeunistd.h: Don't define ssize_t here for MSVC - do it in config.h instead. This avoids pulling in <windows.h> which will hopefully fix another macro redefinition warning under MSVC. Wed May 02 15:20:56 BST 2007 Olly Betts <olly@survex.com> * common/safewindows.h: Define NOGDI before including windows.h to prevent it from defining a macro called "ERROR" which collides with the generated queryparser sources, causing a warning in the MSVC build. Wed May 02 14:38:33 BST 2007 Olly Betts <olly@survex.com> * common/termlist.h: Forward declare ExpandStats as "class" not "struct" for consistency with the actual declaration. Wed May 02 14:30:41 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Try '-lzlib' for zlib, which mingw reportedly needs. Wed May 02 14:21:40 BST 2007 Olly Betts <olly@survex.com> * common/remoteconnection.h,include/xapian/dbfactory.h, net/remoteserver.cc: A timeout of `0' in the remote backend now means "no timeout" (code patch from Mark Hammond). Wed May 02 04:17:28 BST 2007 Olly Betts <olly@survex.com> * Makefile.am,api/,backends/alltermslist.cc,backends/flint/, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/multi/multi_termlist.cc,backends/multi/multi_termlist.h, backends/quartz/,backends/remote/net_termlist.cc, backends/remote/net_termlist.h,backends/remote/remote-database.cc, common/,expand/,matcher/: Split expand functionality out of "matcher" subdirectory into new "expand" subdirectory since it's not really connected to query matching and the matcher subdirectory is one of the largest. Rewrite OrTermList, eliminating BranchTermList entirely. Alter expand to pass in an object (of class Xapian::Internal::ExpandWeight) which accumulates statistics instead of returning OmExpandBits objects which we then have to merge (and eliminating the need for LeafTermList entirely). Query expansion for multiple databases should now give results more like that for a single database with the same documents (if USE_EXACT_TERMFREQ is used, the results should be the same). Debug logging from expand is now all of type EXPAND (some was of types MATCHER and WTCALC before). Wed May 02 03:35:17 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Add deprecation info for Omega too. Wed May 02 00:24:13 BST 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Add test expandweights1 as a regression test for the previous commit. * tests/harness/testsuite.h: Set the precision for TEST_EQUAL_DOUBLE so we don't claim two numbers to be different yet display them the same! Tue May 01 23:27:10 BST 2007 Olly Betts <olly@survex.com> * backends/remote/net_termlist.cc,backends/remote/remote-database.cc, common/remoteprotocol.h,docs/remote_protocol.html, net/remoteserver.cc: Fix handling of the document length in the termlist for the remote backend. Tue May 01 23:17:28 BST 2007 Olly Betts <olly@survex.com> * tests/api_db.cc: Several tests were marked as "local db only" for historical reasons. Enable them for remote backends too. * backends/remote/net_postlist.cc,tests/api_db.cc: Fix a bug in NetworkPostList::skip_to() which apitest's postlist4 reveals. Tue May 01 21:00:26 BST 2007 Olly Betts <olly@survex.com> * api/omtermlistiterator.cc,api/omvalueiterator.cc, backends/multi/multi_alltermslist.cc: Add explicit `#include "omassert.h"'. * backends/multi/multi_alltermslist.cc: Remove all explicit std:: qualifiers and add `using namespace std;'. Tue May 01 20:56:10 BST 2007 Olly Betts <olly@survex.com> * common/multialltermslist.h: Add explicit std:: qualifiers for string, for consistency with std::vector in this header. Tue May 01 15:19:53 BST 2007 Olly Betts <olly@survex.com> * matcher/branchpostlist.h: Don't check for NULL pointer before calling delete. Tue May 01 02:29:10 BST 2007 Olly Betts <olly@survex.com> * common/document.h,common/documentterm.h,common/multialltermslist.h, matcher/rset.cc,queryparser/queryparser.cc: Add some explicit includes for headers which were previously only implicitly included (in preparation for a forthcoming change which removes some of the implicit inclusions). Tue May 01 01:20:53 BST 2007 Olly Betts <olly@survex.com> * backends/database.cc,bin/quartzcheck.cc,docs/deprecation.rst, include/xapian/dbfactory.h,tests/api_db.cc, tests/harness/backendmanager.cc,tests/quartztest.cc: Deprecate the Quartz backend and related functions and utilities. Fri Apr 27 12:56:13 BST 2007 Olly Betts <olly@survex.com> * tests/Makefile.am: Pass '-no-install' when linking test programs, since we don't ever install them. This means libtool doesn't need to generate shell script wrappers on most platforms. Also prefer '$(ldflags)' to '@ldflags' as the former allows the user to override when they run 'make' which is sometimes useful. Fri Apr 27 03:44:15 BST 2007 Olly Betts <olly@survex.com> * docs/intro_ir.html: Add links for the various researchers mentioned by name. Thu Apr 26 17:57:10 BST 2007 Olly Betts <olly@survex.com> * docs/bm25.html,docs/intro_ir.html: Assorted documentation improvements. Thu Apr 26 15:43:24 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Update with changes to Python bindings. Thu Apr 26 06:01:45 BST 2007 Olly Betts <olly@survex.com> * common/expandweight.h,matcher/expandweight.cc: Better handling of expand with multiple databases. Wed Apr 25 18:35:13 BST 2007 Olly Betts <olly@survex.com> * api/omenquire.cc: Delete nonsensical comment. Wed Apr 25 17:41:58 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Use the collective term "features" rather than overloading the term "functions". Casual readers may miss the definition altogether, and even dedicated readers will probably already have a strong mental image of what a "function" is in the context of an API. Wed Apr 25 14:52:11 BST 2007 Olly Betts <olly@survex.com> * docs/intro_ir.html: Improve the look of the formulae. Wed Apr 25 03:56:18 BST 2007 Olly Betts <olly@survex.com> * common/omqueryinternal.h,languages/generate-allsnowballheaders.in, matcher/msetcmp.h,queryparser/queryparser_internal.h: Add missing multiple inclusion guards. Wed Apr 25 03:12:50 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser_internal.h: Don't include <config.h> in a header file. Wed Apr 25 03:05:09 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.cc: Initialise `internal' member inline for consistency with other classes. Tue Apr 24 04:55:45 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Add missing ')' to the error messages when zlib stuff isn't found. Tue Apr 24 02:56:57 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.h, backends/inmemory/inmemory_database.h, backends/quartz/quartz_database.h: MSVC doesn't handle using properly. Tue Apr 24 01:25:34 BST 2007 Olly Betts <olly@survex.com> * matcher/bm25weight.cc,matcher/expandweight.cc: Fix typo in comments: 'releveant' -> 'relevant'. Mon Apr 23 23:00:23 BST 2007 Olly Betts <olly@survex.com> * api/omenquire.cc,docs/deprecation.rst,include/xapian/enquire.h, tests/api_anydb.cc,tests/api_db.cc: Deprecate Enquire::include_query_terms and Enquire::use_exact_termfreq in favour of capitalised versions Enquire::INCLUDE_QUERY_TERMS and Enquire::USE_EXACT_TERMFREQ (for consistency with our other manifest constants, and general C/C++ conventions). Mon Apr 23 18:14:20 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc,backends/quartz/btree.cc: Support for VLAs in C++ is a GCC extension, so eliminate the need for them. Mon Apr 23 16:32:07 BST 2007 Olly Betts <olly@survex.com> * docs/: svn:ignore deprecation.html. svn:ignore apidoc.pdf and sourcedoc.pdf rather than *.pdf. Mon Apr 23 16:08:53 BST 2007 Olly Betts <olly@survex.com> * tests/quartztest.cc: Resolve FIXME about MapTermList. Mon Apr 23 16:06:38 BST 2007 Olly Betts <olly@survex.com> * api/valuerangeproc.cc,backends/flint/flint_check.cc, backends/flint/flint_table.cc,backends/quartz/btree.cc, backends/quartz/btreecheck.cc,bin/xapian-tcpsrv.cc,examples/, languages/steminternal.h,tests/harness/testsuite.cc, tests/harness/testsuite.h,tests/stemtest.cc: Fix compilation with GCC 4.3 snapshot. Mon Apr 23 15:20:55 BST 2007 Olly Betts <olly@survex.com> * common/safeunistd.h: Comment tweaks. Mon Apr 23 11:54:14 BST 2007 Richard Boulton <richard@lemurconsulting.com> * common/safeunistd.h: Include <process.h> on windows to get getpid() instead of using a #define. Mon Apr 23 11:30:56 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/maptermlist.h: Fix typo in error message. Mon Apr 23 11:08:01 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/maptermlist.h: Throw an InvalidOperationError() if MapTermList::get_termfreq() is called, instead of simply having Assert(false): this gets called if a term iterator is accessed for a freshly created Document, and the get_termfreq() method is called on it. The error message is specific to this situation, since this is the only place in which MapTermList is used. Mon Apr 23 03:22:54 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_lock.cc, backends/flint/flint_lock.h: Report why we failed to get a write lock on a flint database. Mon Apr 23 01:44:13 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_version.cc: Update the flint format version since older flint versions can't read compressed tags. Mon Apr 23 01:25:25 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Probe for zlib.h and -lz. * backends/flint/,bin/xapian-compact.cc: Use zlib to compress tags in the record and termlist tables. * tests/Makefile.am,tests/runtest.in,tests/valgrind.supp: Add a valgrind suppression for zlib (a known issue - for speed zlib deliberately reads past the end of the buffer in some cases). Sun Apr 22 04:24:06 BST 2007 Olly Betts <olly@survex.com> * backends/flint/: Remove documentation for blocksize_ parameters which were removed some time ago. Sun Apr 22 00:02:13 BST 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Track the minimum weight required to be considered for the MSet separately from the minimum item which could be considered. Trying to combine the two is causing subtle bugs (fixed bug#86). Sun Apr 22 00:00:39 BST 2007 Olly Betts <olly@survex.com> * HACKING: Add notes about the files Omega uses from xapian-core. Sat Apr 21 23:49:58 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.cc: Clear the stoplist when we parse a new query. * tests/queryparsertest.cc: Add regression test. Sat Apr 21 23:27:33 BST 2007 Olly Betts <olly@survex.com> * Makefile.am,common/Makefile.mk,getopt/: Move getopt.cc from getopt/ to common/. Having a whole directory for a single source file is a bit extravagant, and putting it in common/ allows us to share a copy with omega easily. Sat Apr 21 21:38:20 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/quartz/quartz_database.cc, backends/quartz/quartz_log.cc,common/omdebug.h,common/safeunistd.h, tests/harness/testsuite.cc: safeunistd.h now defines getpid() to GetCurrentProcessId() for MSVC, and we now include this where we want to use getpid() rather than conditionally defining getpid(). Sat Apr 21 04:13:12 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Explicitly set AM_CXXFLAGS to an empty value to start with to avoid issues if it's set in the environment. Cache the results of the test to see if the C++ compiler actually works, and in the XAPIAN_TEST_LINKER_FLAG macro. * configure.ac, include/xapian/visibility.h: Add --disable-visibility option to configure to completely disable use of -fvisibility. * INSTALL: Document all --enable-backend-* options explicitly. Document how to disable use of -fvisibility and -Bsymbolic-functions should you want/need to. Improve wording in a few places. Fri Apr 20 19:20:23 BST 2007 Olly Betts <olly@survex.com> * HACKING,configure.ac: Revert to requiring autoconf 2.59 so that the .spec file for building RPMs can run `autoreconf' to work around a libtool bug. Fri Apr 20 18:12:51 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Document return type of ExpandDecider::Apply() and MatchDecider::Apply() in C# is now bool instead of int. Fri Apr 20 18:01:18 BST 2007 Olly Betts <olly@survex.com> * xapian.spec.in: Package xapian-check and its man page. Fri Apr 20 14:38:00 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Note changed return type of ExpandDecider::operator() and MatchDecider::operator(). Fri Apr 20 14:28:39 BST 2007 Olly Betts <olly@survex.com> * include/xapian/enquire.h: MatchDecider::operator() now returns `bool' not `int'. Remove declaraction of ExpandDecider. * include/xapian/expanddecider.h: Move declaration of ExpandDecider to here. ExpandDecider::operator() now returns `bool' not `int'. ExpandDeciderAnd now names parameters `first' and `second' and documents the order in which they are applied (useful if you have an expensive and a cheap test). ExpandDeciderAnd can now take parameters by reference instead of pointer. * api/Makefile.mk,include/xapian/enquire.h,api/omenquire.cc, api/expanddecider.cc: New home for definitions of virtual methods of ExpandDecider and subclasses. * api/omenquire.cc,common/expand.h,matcher/expand.cc: Eliminate internal class ExpandDeciderAlways - just test pointer against NULL instead. * tests/api_anydb.cc,tests/api_db.cc: Update for change in return type from MatchDecider::operator() and ExpandDecider::operator(). Fri Apr 20 10:35:09 BST 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Add note on rules with multiple targets. Thu Apr 19 19:02:42 BST 2007 Olly Betts <olly@survex.com> * PLATFORMS: Remove reports for 0.7.* and demote reports for 0.8.* to "older reports" status. All SF compilefarm machines are now "no longer available", so update the symbols and key to reflect this. Thu Apr 19 13:27:59 BST 2007 Olly Betts <olly@survex.com> * HACKING: Fix vim magic comment. Add space after URL to stop '::' getting glued to it. Thu Apr 19 11:36:19 BST 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Turn syntax highlighting off, since it's broken for the restructured text in this file anyway. Thu Apr 19 03:26:28 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.h, backends/inmemory/inmemory_database.h, backends/quartz/quartz_database.h: The trick of importing base class methods with 'using' breaks compilation with GCC 2.95 which doesn't issue the warning we're trying to avoid, so add a preprocessor check, Thu Apr 19 03:20:31 BST 2007 Olly Betts <olly@survex.com> * common/progclient.h,net/progclient.cc: Always use pid_t not int for holding a process id. Thu Apr 19 01:08:57 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Update '#callable' footnote. Thu Apr 19 00:07:02 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.h, backends/inmemory/inmemory_database.h, backends/quartz/quartz_database.h: Add using declarations to subclasses of Xapian::Database::Internal which don't override the term forms of delete_document() and replace_document() to prevent compiler warnings about these methods being hidden. Wed Apr 18 23:22:29 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Remove code which can't actually be reached, which fixes a warning in the mingw build. Wed Apr 18 23:07:45 BST 2007 Olly Betts <olly@survex.com> * api/omenquire.cc,common/multimatch.h,common/omenquireinternal.h, include/xapian/enquire.h,matcher/Makefile.mk, matcher/biaspostlist.h,matcher/multimatch.cc,net/remoteserver.cc: Remove Enquire::set_bias(). * docs/deprecation.rst: Update wrt Enquire::set_bias(). Fix typo. Wed Apr 18 18:51:42 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Require autoconf 2.60 or newer. Drop docdir compatibility hack which is no longer required. * HACKING: Document requirement. Fix typo (or out-of-date info) for automake requirement (we require 1.9.5 not 1.8.5). Wed Apr 18 18:27:35 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Note the deprecation of get_description() methods for Python. Wed Apr 18 14:57:03 BST 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Minimal changes to make this into valid restructured text. Wed Apr 18 11:27:57 BST 2007 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Require autoconf version 2.59c or later: AC_TYPE_SSIZE_T isn't present in earlier versions. Wed Apr 18 11:08:13 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Fix table formatting error Wed Apr 18 01:32:48 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Update with methods which have been removed from the bindings. Wed Apr 18 00:24:00 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Move methods which have just been removed from the list of deprecated methods to the list of removed methods. Tue Apr 17 23:58:53 BST 2007 Olly Betts <olly@survex.com> * api/omdocument.cc,api/omenquire.cc,api/omquery.cc,include/xapian/, languages/stem.cc,tests/api_db.cc: Remove methods we're deprecating for 1.0. Tue Apr 17 23:50:06 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Note that is_empty() is not deprecated for PHP after all! Tue Apr 17 22:17:41 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: GCC 2.95 doesn't support std::string::clear(), so use resize(0) instead. * HACKING: Document the above issue in a reworked section on C++ STL features. Update the section on safeXXX.h headers. Tue Apr 17 22:07:39 BST 2007 Olly Betts <olly@survex.com> * common/safesysselect.h: Correct typo in #error message. Tue Apr 17 19:18:00 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Factor out code to test for a linker flag into macro XAPIAN_TEST_LINKER_FLAG. With g++, use XAPIAN_TEST_LINKER_FLAG to see if -Bsymbolic-functions is supported (it requires a very recent version of ld currently). This option reduces the size and load time of the shared library by resolving references within the library when it's created. Currently untested on a box which supports -Bsymbolic-functions. Tue Apr 17 12:01:46 BST 2007 Olly Betts <olly@survex.com> * unicode/utf8itor.cc: Tidy up code layout. Tue Apr 17 02:06:39 BST 2007 Olly Betts <olly@survex.com> * include/xapian/unicode.h,unicode/utf8itor.cc: Make nonascii_to_utf8 a public method. Make get_case_type(), get_category(), and get_delta() internal, but provide a public get_category() function which takes a Unicode character value. Tue Apr 17 02:05:32 BST 2007 Olly Betts <olly@survex.com> * backends/flint/Makefile.mk: Distribute flint_check.h. Tue Apr 17 01:47:05 BST 2007 Olly Betts <olly@survex.com> * tests/api_posdb.cc: The remote backend now supports all database operations, so merge localpositionaldb_tests into positionaldb_tests. * tests/: Remove the trivial separate api_XXXdb.h headers, some of which were out-of-step with their respective .cc sources. Instead just keep the array external declarations in apitest.h. * tests/apitest.h,tests/api_transdb.cc: Move TESTCASE and END_OF_TESTCASES macros into header. * tests/api_db.cc: Remove test collapsekey2 - it's only relevant for the Muscat 3.6 backend, so it's unused now that has been removed. * tests/: Add unicode tests, adapted from Omega's utftest. Tue Apr 17 00:14:16 BST 2007 Olly Betts <olly@survex.com> * Makefile.am,include/Makefile.mk,include/xapian.h, include/xapian/unicode.h,queryparser/,unicode/: Expose Unicode and UTF-8 related classes and functions as a public API. Mon Apr 16 16:06:22 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Bring document fully up-to-date, listing all the planned deprecations for the bindings. Also, include the full upgrade notes in the listings, so they can be preserved accessibly for posterity. I even retrieved the notes for the already-removed QueryParser::set_stemming_options from 0.9.10. Mon Apr 16 11:36:18 BST 2007 Olly Betts <olly@survex.com> * net/tcpclient.cc: Call WSAGetLastError() instead of socket_errno() when we want to compare the result against WSAEWOULDBLOCK. Mon Apr 16 11:19:21 BST 2007 Olly Betts <olly@survex.com> * common/remoteconnection.h,common/safewinsock2.h: Need to negate the POSIX error codes we define in terms of winsock ones. Fri Apr 13 16:24:29 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/doxygen_api_conf.in: Turn on XML output, so that we can use it to populate docstrings for the python bindings. Fri Apr 13 11:38:39 BST 2007 Olly Betts <olly@survex.com> * backends/Makefile.mk,backends/alltermslist.cc,common/alltermslist.h: Split AllTermsList into header and code. AllTermsList::get_wdf() now throws InvalidOperationError() rather than asserting in a debug build and returning 0 in a non-debug build. * backends/Makefile.mk: Correct list for "how to add a new backend". Fri Apr 13 10:17:46 BST 2007 Olly Betts <olly@survex.com> * api/version.cc: Simpler and cleaner fix - the prototypes for the version functions had moved from xapian/version.h to xapian.h so we were including the wrong header and not seeing the prototypes with visibility markup! Fri Apr 13 09:29:17 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/version.cc: Add visibility annotations to definitions of version functions: the annotation apparently needs to be applied to the definition as well as the declaration or it is ignored: possibly the annotation on the declaration is ignored anyway, but it should be kept there for documentation purposes in any case. Fixes link error with bindings under gcc 4.1. Fri Apr 13 01:39:57 BST 2007 Olly Betts <olly@survex.com> * common/remoteconnection.h,net/progclient.cc,net/remoteconnection.cc: GetLastError() has an unsigned return type, so need to cast to int before negating. Fri Apr 13 01:39:20 BST 2007 Olly Betts <olly@survex.com> * docs/deprecation.rst: Add xapian_version_string() and friends. Fri Apr 13 01:24:53 BST 2007 Olly Betts <olly@survex.com> * tests/api_nodb.cc: More 'tout' output. Fri Apr 13 01:19:20 BST 2007 Olly Betts <olly@survex.com> * api/Makefile.mk,api/error.cc,bin/xapian-tcpsrv.cc, common/remoteconnection.h,common/safeerrno.h,configure.ac, generate-exceptions.in,net/,tests/harness/testsuite.cc: On MS Windows, if Xapian::Error::my_errno is the result of GetLastError() or WSAGetLastError(), negate it (the error codes can overlap with errno codes). On UNIX, put h_errno negated in my_errno if a call to gethostbyname() fails. Add Xapian::Error::get_error_string() method which decodes my_errno in the appropriate way to give a string, and use this instead of strerror(error.get_errno()). Fri Apr 13 01:17:08 BST 2007 Olly Betts <olly@survex.com> * languages/generate-allsnowballheaders.in: win32 makefiles don't keep the snowball header list in sorted order like we do, so fix that here. Also allow for '\' as path separator. Fri Apr 13 00:58:09 BST 2007 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Add "tout" debug output to stemlangs1. Thu Apr 12 21:37:04 BST 2007 Olly Betts <olly@survex.com> * tests/Makefile.am: Add support for QUIET= and QUIET=y arguments to "make". Thu Apr 12 21:35:50 BST 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Use a substitution for RSTHTML (like we do for the snowball generated sources). Thu Apr 12 20:41:18 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Add validate_query() calls to the "term" and "value range" constructors. Thu Apr 12 20:00:14 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,include/xapian/query.h: Only validate a Query object when it's either constructed or changed to avoid O(n^2) behaviour in some cases. Thu Apr 12 19:59:13 BST 2007 Olly Betts <olly@survex.com> * queryparser/lemon.c: Fix nasty function casting for argument to qsort - it's (at least technically) not portable. Thu Apr 12 17:47:47 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Fix GCC version test to only turn on -fvisibility=hidden for GCC >= 4. Thu Apr 12 16:47:56 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_btreebase.h,backends/flint/flint_cursor.h, backends/flint/flint_table.h,backends/quartz/,common/,configure.ac, generate-exceptions.in,include/Makefile.mk,include/xapian.h, include/xapian/: Add visibility annotations to the library, which when using GCC >= 4 reduces the size and load time of the library and increase the runtime speed a little. Under x86_64, the stripped library is 6.4% smaller (1.5% smaller with debug information). Thu Apr 12 16:45:57 BST 2007 Olly Betts <olly@survex.com> * api/version.cc: Add missing "#include <config.h>". Thu Apr 12 13:41:52 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/quartztest.cc: Put removedir() function back for now to fix build. Thu Apr 12 13:02:30 BST 2007 Richard Boulton <richard@lemurconsulting.com> * net/progclient.cc: Some (but not all) versions of MSVC need "#include <io.h>" to get the _open_osfhandle identifier. Add it. Thu Apr 12 12:00:34 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/remote/net_postlist.cc: Include config.h Thu Apr 12 11:54:50 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/multi/multi_termlist.cc: Include config.h to get system specific defines. * common/safesysstat.h: Change signature of mkdir to take mode_t instead of int, to match unixy definition. Thu Apr 12 10:19:49 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Probe for mode_t. Thu Apr 12 01:10:18 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Move BitReader class inside Xapian namespace to avoid namespace pollution. * bin/xapian-check.cc: Include BitReader::decode_interpolative() - previously we were relying on using the copy from the library. Thu Apr 12 01:02:53 BST 2007 Olly Betts <olly@survex.com> * common/safesysstat.h: Under __WIN32__, provide a POSIX-like mkdir() wrapper which takes 2 arguments, but ignores the file mode, so we can just call mkdir with a mode argument everywhere. * common/utils.h: Remove special case handling of mkdir. * bin/xapian-compact.cc: Replace uses of om_tostring() with a call to sprintf. No longer need to `#include "utils.h"'. Thu Apr 12 00:44:18 BST 2007 Olly Betts <olly@survex.com> * api/,backends/inmemory/inmemory_positionlist.cc,common/,matcher/, net/remoteconnection.cc,net/serialise.cc: omdebug.h only includes omassert.h for the trivial "STRINGIZE" macro, so just duplicate that rather than pulling in an extra header everywhere. Then fix up the two dozen files which were relying on this implicit inclusion! Thu Apr 12 00:10:48 BST 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Tweak #ifdef XAPIAN_DEBUG_VERBOSE to include DEBUGLINE() for clarity (no change in behaviour). Wed Apr 11 23:36:10 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix regression introduced by Richard's change, and restructure to make the handling for each case more obvious. * tests/queryparsertest.cc: Add a couple of testcases for the regression. * queryparser/queryparser_internal.h: Remove reference to "bool" in comment which no longer applies. Wed Apr 11 21:25:36 BST 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser_internal.h: Refactor boolean "flag" member in BoolAndString class into an enum of possible prefix types. Rename "flag" to "type". Rename BoolAndString to PrefixInfo. * queryparser/queryparser.cc: Change add_prefix() and add_boolean_prefix() to match changes to queryparser_internal.h * queryparser/queryparser.lemony: Refactor handling of prefixed terms to match queryparser_internal, and to use a switch() structure: no functional changes, but code is considerably more readable (I hope). Wed Apr 11 21:13:31 BST 2007 Olly Betts <olly@survex.com> * tests/api_db.cc,tests/quartztest.cc: Rework quartztest's "open1" into API level tests for quartz and flint in apitest. Wed Apr 11 14:01:31 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/deprecation.rst: Remove documentation of the "#define XAPIAN_DEPRECATED(D) D" hack, and clarify documentation on how long we support deprecated functions. Wed Apr 11 09:25:55 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/Makefile.am,docs/deprecation.rst: Add initial draft of deprecation policy document, with list of all API items currently marked for deprecation. * HACKING: Add note about keeping deprecation.rst up-to-date when making a release. Tue Apr 10 19:07:24 BST 2007 Olly Betts <olly@survex.com> * configure.ac: --enable-quiet now uses AS_HELP_STRING. Tue Apr 10 18:53:32 BST 2007 Olly Betts <olly@survex.com> * tests/harness/index_utils.cc: Return the empty document we already have rather than constructing another. Tue Apr 10 17:45:11 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Fix typo in comment. Tue Apr 10 17:35:57 BST 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Fix "if" to "is" typo in 3 comments. Tue Apr 10 16:46:09 BST 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Define XAPIAN_TCPSRV and XAPIAN_PROGSRV under UNIX too, and use them everywhere. Tue Apr 10 16:33:48 BST 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.h: Remove "BackendManager::" from method declaration. Tue Apr 10 15:35:50 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Define HAVE_SOCKETPAIR if we have socketpair(). Tue Apr 10 15:29:49 BST 2007 Olly Betts <olly@survex.com> * net/progclient.cc: Fix #endif placement so it compiles on UNIX. Tue Apr 10 14:18:06 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Remove debugging code. Tue Apr 10 12:57:51 BST 2007 Olly Betts <olly@survex.com> * backends/dbfactory_remote.cc,common/progclient.h,net/progclient.cc, net/remoteconnection.cc,tests/api_db.cc,tests/apitest.cc, tests/harness/backendmanager.cc,tests/harness/backendmanager.h: Add support for the "prog" variant of the remote backend (patch started by me, finished by Mark Hammond). Tue Apr 10 12:23:46 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,backends/inmemory/inmemory_database.cc, backends/quartz/quartz_alldocspostlist.cc,matcher/localmatch.cc, matcher/multimatch.cc,tests/harness/testutils.cc: Prefer `CONTAINER.empty()' to `CONTAINER.size() == 0'. Tue Apr 10 12:09:45 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Tighten up an assertion. Mon Apr 09 20:56:23 BST 2007 Olly Betts <olly@survex.com> * configure.ac,tests/harness/backendmanager.cc,tests/runsrv.in, tests/runtest.in: Run the remote backend server using new "runsrv" script instead of "runtest". This doesn't echo anything to stdout, and if valgrind is in use, runs the remote server under "--tool=none" which is much quicker than using valgrind's default memcheck tool (we need to run the remote server under valgrind because valgrind's emulation of excess FP precision isn't exact and otherwise we get tests failing because of these differences). Mon Apr 09 20:24:30 BST 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc,tests/api_nodb.cc: Move test poscollapse2 to the "no database" category, since it doesn't require a database! Mon Apr 09 18:04:40 BST 2007 Olly Betts <olly@survex.com> * docs/remote_protocol.html: Correct documentation - MSG_REOPEN gives a response of REPLY_UPDATE. Mon Apr 09 15:08:44 BST 2007 Olly Betts <olly@survex.com> * ./: svn:eol-style not svn:eolstyle. Mon Apr 09 15:00:30 BST 2007 Olly Betts <olly@survex.com> * ./: Set svn:eolstyle to native for most text files. Mon Apr 09 14:50:40 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Delete the corresponding entry (if any) from doclens in delete_document(). Add assertion to add_document_() that the corresponding entry in doclens isn't already set, but in a non-debug build overwrite any existing entry as that's more likely to be correct. * backends/quartz/quartz_database.cc: Ditto. Mon Apr 09 14:08:55 BST 2007 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Just assign to doclens[did]. Mon Apr 09 14:06:58 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Just assign to doclens[did]. Mon Apr 09 11:29:22 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/quartz/quartz_database.cc: Apply same change to quartz. Mon Apr 09 11:24:31 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_database.cc: Fix bug with document lengths in replace document. Was using doclens.insert() to set the new document length, but this has no effect if the entry already exists. This fixes replacedoc3 for flint (and thus for remote databases), but the same change is needed for quartz. * backends/flint/flint_postlist.cc: Add some debugging lines. Mon Apr 09 01:43:02 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/: Add some more debugging messages to help track down the problem with document lengths shown up by previous change. Mon Apr 09 01:40:34 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_wrdb.cc: Add asserts to check that the value of get_doclength() on the database and on the posting list are correct. Currently fails with the value from the posting list for the document which has been replaced (or sooner if assertions are turned on). Sun Apr 08 23:24:20 BST 2007 Olly Betts <olly@survex.com> * Makefile.am,configure.ac: `./configure --enable-quiet' allows you to specify at configure time to pass `--quiet' to libtool. Now you can override this at make-time by using `make QUIET=' (to turn off `--quiet') or `make QUIET=y' (to turn on `--quiet'). Sun Apr 08 23:23:50 BST 2007 Olly Betts <olly@survex.com> * HACKING: Mention AssertEqParanoid and AssertNeParanoid. Sun Apr 08 22:45:45 BST 2007 Olly Betts <olly@survex.com> * common/omassert.h: Add AssertEqParanoid and AssertNeParanoid macros. * backends/flint/flint_table.cc,backends/inmemory/inmemory_database.cc, backends/multi/multi_postlist.cc,backends/quartz/btree.cc: Use them. * backends/multi/multi_postlist.cc: Add extra assertions to MultiPostList::get_doclength(). Sun Apr 08 22:16:11 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Fix typos: `AC_CHECK_SSIZE_T' -> `AC_TYPE_SSIZE_T'; `AC_CHECK_PID_T' -> `AC_TYPE_PID_T'. Sun Apr 08 22:04:54 BST 2007 Olly Betts <olly@survex.com> * HACKING: Snapshots and releases are now generated with autoconf 2.61. Sun Apr 08 21:42:03 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Use AC_CHECK_SSIZE_T instead of our own test. Add AC_CHECK_PID_T for any platforms which don't have `pid_t'. * net/tcpserver.cc: Use `pid_t' instead of `int'. * common/progclient.h,net/progclient.cc: Cleaner rewritten version of progclient.h. Sun Apr 08 19:59:43 BST 2007 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc: If kill() fails, don't bother calling waitpid(). Sun Apr 08 19:41:11 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Don't explicitly pass "0" for errno. Sun Apr 08 16:58:58 BST 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.lt: Fix warnings in debugging builds on gcc 4.1 due to comparison of signed and unsigned types. Also, use the ParseTokenName function instead of explicitly doing a range check each time an entry is looked up in yyTokenName, and add a ParseRuleName function for looking up things in yyRuleName, and use it similarly. Sun Apr 08 15:55:42 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Fix lingering references to enable_debug. Sun Apr 08 15:48:13 BST 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc,tests/api_wrdb.cc: Remove hacks to automatically SKIP tests which failed because of lack of support for some features by the remote backend. Sun Apr 08 15:14:11 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,queryparser/queryparser.lt: More tweaks. Sun Apr 08 15:04:49 BST 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.lt: Always include "omdebug.h", and put one of the pieces of debugging code in #ifdef XAPIAN_DEBUG_VERBOSE, to fix non-logging message builds. Sun Apr 08 14:23:06 BST 2007 Richard Boulton <richard@lemurconsulting.com> * docs/remote_protocol.html: Update for protocol version 27 by adding documentation for MSG_POSTLIST and replies. Also, fix a typo. Sun Apr 08 13:50:20 BST 2007 Olly Betts <olly@survex.com> * common/omdebug.h,queryparser/queryparser.lemony, queryparser/queryparser.lt: Hook the debug tracing in the lemon generated parser into Xapian's debug logging framework. Sun Apr 08 12:35:29 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/remote/,common/remote-database.h,common/remoteprotocol.h, common/remoteserver.h,net/remoteserver.cc: Implement postlists for the remote database. Increases XAPIAN_REMOTE_PROTOCOL_VERSION to 27. Sun Apr 08 12:33:02 BST 2007 Olly Betts <olly@survex.com> * HACKING,INSTALL,configure.ac: Rename --enable-debug* - conflating the options to "turn on assertions" and "turn on logging" still confuses me to the extent I need to read the documentation every time I use them. `--enable-debug[=partial]' become `--enable-assertions'; `--enable-debug-verbose' becomes `--enable-log' and `--enable-debug=full' becomes `--enable-assertions --enable-log'. For now the old options give an error telling you the new equivalent. Sun Apr 08 10:45:40 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Eliminate use_quiet - just use enable_quiet instead. Fri Apr 06 19:06:07 BST 2007 Olly Betts <olly@survex.com> * net/progclient.cc: Close stderr of the spawned backend program. Fri Apr 06 18:59:24 BST 2007 Olly Betts <olly@survex.com> * net/progclient.cc: msecs_timeout_ -> msecs_timeout. Fri Apr 06 18:12:36 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Fix comment grammar. Fri Apr 06 18:10:31 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Deal with the exceptional case first. Fri Apr 06 17:51:20 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Use string::append(ptr, size) in preference to string::append(ptr, ptr + size). Fri Apr 06 11:45:39 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,backends/flint/flint_database.cc, backends/flint/flint_database.h,backends/flint/flint_postlist.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/quartz/quartz_database.cc,backends/quartz/quartz_database.h, backends/quartz/quartz_postlist.cc, backends/remote/remote-database.cc,common/database.h, common/remote-database.h: Refactor to eliminate do_open_post_list and any unnecessary call to term_exists(). Fri Apr 06 11:44:25 BST 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Just use TEST on a boolean value rather than TEST_EQUAL against true/false. Fri Apr 06 08:53:46 BST 2007 Olly Betts <olly@survex.com> * net/remoteserver.cc: Remove "N" which we calculate but never use! Fri Apr 06 08:17:00 BST 2007 Olly Betts <olly@survex.com> * backends/flint/Makefile.mk,backends/flint/flint_check.cc, backends/flint/flint_check.h,bin/Makefile.mk,bin/xapian-check.cc: Fix xapian-check not to use libquartzcheck.la. Fri Apr 06 07:20:47 BST 2007 Olly Betts <olly@survex.com> * bin/Makefile.mk,bin/xapian-check.cc: Add new "xapian-check" program which performs consistency checks on a flint database. Fri Apr 06 06:25:36 BST 2007 Olly Betts <olly@survex.com> * bin/quartzcheck.cc: Test if this is a quartz database by looking at "meta" not "record_DB". If "record_DB" is >= 2GB and we don't have a LFS aware stat function then stat can fail even though the file is there. Also open the database explicitly as a Quartz database for extra robustness. Fri Apr 06 06:23:33 BST 2007 Olly Betts <olly@survex.com> * AUTHORS: Thank Daniel Ménard. Thu Apr 05 18:20:53 BST 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: No need to explicitly initialise std::string to ""; not doing so results in a smaller object file. Thu Apr 05 18:20:12 BST 2007 Olly Betts <olly@survex.com> * common/remoteprotocol.h: Note other changes in protocol version 26. Thu Apr 05 17:34:10 BST 2007 Olly Betts <olly@survex.com> * tests/runtest.in: Pass "--leak-resolution=high" to valgrind to prevent unrelated leak reports related to STL classes from being combined. Thu Apr 05 17:26:39 BST 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Remove stray line left over from debugging. Thu Apr 05 16:47:36 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/internaltest.cc: Fix test_serialiselength2: implementation was correct, but test was failing. Needed special casing for n=0 and 1, and was using arguments to string::append(size_type, charT) the wrong way round (stupid C++ API). Also, make it check the return value of decode_length(), since we know what the answer ought to be. Thu Apr 05 14:26:30 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,backends/remote/remote-database.cc, common/serialise.h,net/remoteserver.cc,net/serialise.cc, tests/internaltest.cc: Add a flag parameter to decode_length() to indicate if the decoded value indicates the length of a string in the protocol stream, and if it does, check it against p_end - p after the length has been decoded (fixes bug#117). Thu Apr 05 14:08:02 BST 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: In test uniqueterm1, check the document length after every delete or replace for better coverage. Thu Apr 05 12:41:04 BST 2007 Olly Betts <olly@survex.com> * net/serialise.cc: Encode deltas between docids (rather than the docids themselves) in the RSet serialisation which reduces the size of the encoding if a lot of large docids with small gaps between them are present. * backends/remote/remote-database.cc,net/remoteserver.cc: Encode deltas between termpositions in REPLY_POSITIONLIST, for similar reasons. No need to encode the term length for MSG_DELETEDOCUMENTTERM as it's the last (indeed only) item, and more consistent not to. * common/remoteprotocol.h: Bump protocol version to 26. * docs/remote_protocol.html: Document the new messages MSG_DELETEDOCUMENTTERM and MSG_REPLACEDOCUMENTTERM. Correct reference to REPLY_DOCUMENT to REPLY_DOCDATA. Fix documentation for MSG_QUERY to include the serialised Weight and RSet objects. Fix "copy and paste" error in the description of Flush (should of course be MSG_FLUSH not MSG_CANCEL). Update for changes above. Thu Apr 05 12:18:19 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/database.cc: Keep a RefCntPtr to the LeafPostList used in implementations of delete_document and replace_document with a unique term, so that the postlist isn't leaked. Thu Apr 05 11:56:57 BST 2007 Olly Betts <olly@survex.com> * backends/database.cc: Database::Internal can't call the PostingIterator(PostingIterator::Internal*) ctor (at least under g++ 3.3.5) because it isn't a friend (only class Database is). Can't seem to forward define Database::Internal to make Database::Internal a friend so just use LeafPostList directly as that seems less bad than pulling in the whole of database.h or making PostingIterator::internal public. Thu Apr 05 08:27:26 BST 2007 Olly Betts <olly@survex.com> * HACKING: Need to run svn-tag-release on ixion. Thu Apr 05 01:50:10 BST 2007 Richard Boulton <richard@lemurconsulting.com> * backends/remote/remote-database.cc,common/remote-database.h, common/remoteprotocol.h,common/remoteserver.h,net/remoteserver.cc: Implement unique term variants of delete and replace document for the remote database. Involves adding two new message types to the protocol (and therefore bumping the protocol version to 25), and overriding the standard implementations of delete_document(term) and replace_document(term) in RemoteDatabase. * tests/api_wrdb.cc: Correct text in SKIP_TEST() message. Thu Apr 05 01:04:19 BST 2007 Richard Boulton <richard@lemurconsulting.com> * api/omdatabase.cc,backends/database.cc,common/database.h: Move implementation of delete_document and replace_document with a unique_term from WritableDatabase to Xapian::Database::Internal. This will allow it to be overridden for databases which need a special implementation (such as the remote database). Tue Apr 03 14:30:45 BST 2007 Olly Betts <olly@survex.com> * common/remote-database.h: Remove unused and undefined method RemoteDatabase::get_spawned_socket(). Tue Apr 03 09:18:51 BST 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Under __WIN32__, launch xapian-tcpsrv as a child process communicating via a pipe and handle the port being already in use by trying the next port up (as we do under UNIX). Tue Apr 03 06:50:48 BST 2007 Olly Betts <olly@survex.com> * net/tcpserver.cc: For __WIN32__ and __CYGWIN__, use SO_EXCLUSIVEADDRUSE (if available) on on listening sockets for xapian-tcpsrv. Tue Apr 03 04:37:02 BST 2007 Olly Betts <olly@survex.com> * include/xapian/positioniterator.h,include/xapian/postingiterator.h: Fix mangled comment start in (C) headers. Tue Apr 03 03:27:12 BST 2007 Olly Betts <olly@survex.com> * include/xapian/,generate-exceptions.in: Mark the following constructors 'explicit': DocIDWrapper(docid), TermPosWrapper(const std::string &), Query::Internal(const std::string &), Database(const std::string &), all Error subclass constructors taking const std::string &. Only the Database and Error subclass changes should be able to affect user code at all. * include/xapian/,api/omenquire.cc: Make the following methods 'const': RSet::contains(MSetIterator), Enquire::get_query(). This shouldn't affect the validity of any user code. Mon Apr 02 11:44:26 BST 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Add note on "Building from SVN on Windows with MSVC" Mon Apr 02 11:32:51 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Fix assertion to allow for OP_VALUE_RANGE. Mon Apr 02 11:14:10 BST 2007 Olly Betts <olly@survex.com> * autoconf/rjb_find_stlport.m4: Fix check that the user didn't say `./configure --with-stlport-compiler' (without `=gcc' or similar after it). Mon Apr 02 10:38:35 BST 2007 Olly Betts <olly@survex.com> * tests/api_db.cc,tests/apitest.cc: Under __WIN32__ disable tests which require the prog variant of the remote backend. Mon Apr 02 09:44:13 BST 2007 Olly Betts <olly@survex.com> * net/tcpserver.cc: Don't use SO_REUSEADDR under __CYGWIN__ or __WIN32__ as it has incorrect semantics. Mon Apr 02 09:37:14 BST 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Pull `1239' out as `DEFAULT_PORT'. On Windows, cycle through 10 ports starting at DEFAULT_PORT so that we can have multiple databases open at once. Add a monotonic count to the name of the log file created from the output of each xapian-tcpsrv to avoid failures because the log file is already open. Mon Apr 02 07:41:57 BST 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Avoid having the same database open twice at once. Xapian itself copes, but the BackendManager class in the test harness struggles in the remotetcp case under mingw. Mon Apr 02 07:14:13 BST 2007 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Make Enquire constructor explicit since it can take a single parameter (of type Database). Mon Apr 02 02:32:23 BST 2007 Olly Betts <olly@survex.com> * net/tcpserver.cc: Add "#error" if neither HAVE_FORK nor __WIN32__ is defined (otherwise we'd failed with a rather obscure link error much later!) Mon Apr 02 01:54:20 BST 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Remove unused variable left over from recent change. Mon Apr 02 01:22:56 BST 2007 Olly Betts <olly@survex.com> * net/tcpserver.cc: Actually remove SOCKOPT_OPTIONS_TYPE. Mon Apr 02 01:16:03 BST 2007 Olly Betts <olly@survex.com> * net/tcpclient.cc,net/tcpserver.cc: Pass the 4th parameter of setsockopt() as char* which works whether the function actually takes char* or void* (since C++ allows implicit conversion from char* to void*). Only call socket_errno() after gethostbyname() under __WIN32__ - on UNIX the error code is in h_errno (with incompatible values to errno) and we don't want to confuse things by reporting some random value of errno. Mon Apr 02 00:44:56 BST 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Factor out localhost address into LOCALHOST. Use new xapian-tcpsrv --interface option to only listen for connections on localhost. Run xapian-tcpsrv with "start /B" on MS Windows (this stops the "flickbook of console windows" effect). Mon Apr 02 00:41:31 BST 2007 Olly Betts <olly@survex.com> * common/Makefile.mk,common/safesyssocket.h: Oops, I checked in safesysselect.h as safesyssocket.h! Mon Apr 02 00:30:54 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Fix SimpleStopper::add to take `const std::string &' not `const std::string'. Mon Apr 02 00:20:45 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-tcpsrv.cc,net/tcpserver.cc: Add "--interface" option to allow the hostname or address of the interface to listen on to be specified (default is the previous behaviour of listening on all interfaces). * net/remoteconnection.cc,net/tcpclient.cc,common/safesysselect.h: Factor out portability code for sys/select.h. * common/Makefile.mk: Ship common/safesysselect.h. * net/tcpserver.cc: Rename parameter of run_thread from "_param" to "param_" (identifiers with leading underscores are reserved for the compiler). * common/tcpclient.h,net/tcpclient.cc,common/tcpserver.h: Rewrite headers with comments which are actually accurate and up-to-date! TcpClient ctor is small so put inline in the header. Pass std::string by const reference. * common/tcpclient.h,common/tcpserver.h: TcpClient and TcpServer classes now inherit from WinsockInitializer under __WIN32__ (as a mixin class) instead of having a member of type WinsockInitializer (the member is initialised too late, at least under mingw). Sun Apr 01 19:14:11 BST 2007 Olly Betts <olly@survex.com> * common/omdebug.cc: Fix retrying of write to retry the part of the string not already written! Sun Apr 01 18:37:15 BST 2007 Olly Betts <olly@survex.com> * common/remoteconnection.h: Note in comments about use as a "mixin" class. Sun Apr 01 17:42:19 BST 2007 Olly Betts <olly@survex.com> * tests,examples,bin: Add .exe versions of built executables to svn:ignore. Sun Apr 01 16:28:26 BST 2007 Olly Betts <olly@survex.com> * common/omdebug.cc: Remove unused '#include <stdio.h>' - we switched to using unbuffered I/O for logging long ago! Sun Apr 01 15:21:04 BST 2007 Olly Betts <olly@survex.com> * common/safewinsock2.h: No, don't typedef socklen_t here - we use SOCKLEN_T everywhere and have `#define SOCKLEN_T int' in config.h.win32. Sun Apr 01 13:32:21 BST 2007 Richard Boulton <richard@lemurconsulting.com> * common/omdebug.cc: Include necessary headers to get O_CREAT and friends defined (according to documentation in linux manpage), to fix compile error. Also, don't ignore the return value of write() to fix compile warning (and to ensure that the whole output gets written if possible). Sun Apr 01 13:19:52 BST 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.cc: Fix Assert() on a RefCountPtr to check the value of the pointer, to fix compile with asserts turned on. Sun Apr 01 12:57:06 BST 2007 Richard Boulton <richard@lemurconsulting.com> * common/safewinsock2.h: Define socklen_t under MS compiler (since Olly reported build success with mingw, presumably it's not needed there). Sun Apr 01 06:42:14 BST 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk: $(snowball_algorithms:.sbl=.cc) is portable according to the automake list so use that so we only need to list the .sbl sources. Sun Apr 01 05:53:29 BST 2007 Olly Betts <olly@survex.com> * common/tcpserver.h,net/tcpserver.cc: TcpServer's port member variable is set but never used so remove it. Sun Apr 01 03:27:30 BST 2007 Olly Betts <olly@survex.com> * bin/Makefile.mk,examples/Makefile.mk: Add $(EXEEXT) to dependency on binaries in man page generation rules so they work on Windows. Sat Mar 31 22:17:55 BST 2007 Olly Betts <olly@survex.com> * common/multimatch.h,matcher/multimatch.cc: When using a MatchDecider with remote database(s), don't rerun the MatchDecider on documents which a remote server has already checked. Sat Mar 31 10:44:26 BST 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Sort out path to xapian-tcpsrv under mingw. If neither HAVE_FORK nor __WIN32__ is defined, make compilation fail with a suitable #error. Sat Mar 31 10:17:06 BST 2007 Olly Betts <olly@survex.com> * api/Makefile,backends/Makefile,backends/flint/Makefile, backends/inmemory/Makefile,backends/multi/Makefile, backends/quartz/Makefile,backends/remote/Makefile,bin/Makefile, common/Makefile,examples/Makefile,getopt/Makefile,include/Makefile, languages/Makefile,matcher/Makefile,net/Makefile, queryparser/Makefile,tests/harness/Makefile: Don't pass $(MAKEFLAGS) on the command line to $(MAKE) when invoking the top level Makefile from a static Makefile. It's passed automatically anyway, and also $(MAKEFLAGS) doesn't have a leading '-' so "make -s" tries to do "make s" at the top level. Sat Mar 31 09:28:06 BST 2007 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Fix end of loop handling. Sat Mar 31 09:03:26 BST 2007 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Fix to compile. Sat Mar 31 07:53:25 BST 2007 Olly Betts <olly@survex.com> * common/safewindows.h,common/safewinsock2.h: safewindows.h is included by a lot of files so we want to keep it lightweight so split winsock2.h related stuff into safewinsock2.h. * common/remoteconnection.h: Move WinsockInitializer here. * common/safeerrno.h: #include "safewinsock2.h" not <winsock2.h>. Sat Mar 31 07:19:06 BST 2007 Olly Betts <olly@survex.com> * common/remoteserver.h: Stop trying to include "remoteserver.h" because that's the SAME HEADER! Sat Mar 31 07:16:23 BST 2007 Olly Betts <olly@survex.com> * common/tcpclient.h,common/tcpserver.h: Fix comments - the class is called WinsockInitializer not WinsockInitialiser. Sat Mar 31 06:54:22 BST 2007 Olly Betts <olly@survex.com> * net/progclient.cc: Don't compile any code from this file if __WIN32__ is defined (for mingw). Sat Mar 31 06:52:27 BST 2007 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Simplify code in test_stemlangs1(). Sat Mar 31 06:22:26 BST 2007 Olly Betts <olly@survex.com> * net/remoteconnection.cc: Use const_cast to avoid GCC warning about casting away const. Use string::data() rather than string::c_str() since we don't need a terminating zero byte. Sat Mar 31 06:20:07 BST 2007 Olly Betts <olly@survex.com> * backends/dbfactory_remote.cc: Suppress "unused parameter" warnings. Sat Mar 31 06:17:35 BST 2007 Olly Betts <olly@survex.com> * net/tcpserver.cc: Make global variable and function used only in this file static. Don't call delete on a void* (that's undefined behaviour) - delete the cast version of the pointer instead. Sat Mar 31 06:17:20 BST 2007 Olly Betts <olly@survex.com> * AUTHORS: Add a number of bug reporters I missed before. Sat Mar 31 04:49:27 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Clean up handling of --enable-backend-* options. Let mingw try to build the remote backend. Sat Mar 31 03:47:49 BST 2007 Olly Betts <olly@survex.com> * tests/internaltest.cc: Don't compile test_serialisequery1 if the remote backend is disabled to avoid an "unused function" warning (we were already disabling the use of it). Fri Mar 30 22:43:27 BST 2007 Olly Betts <olly@survex.com> * HACKING: Correct several inaccuracies: --enable-debug and --enable-debug=partial don't cause the build system to compile in debugging symbols (configure defaults to adding "-g" for GCC, while for other compilers you must enable them by hand). Also, --enable-debug doesn't "produce warnings", but causes Xapian::AssertionError to be thrown. Add explanation for why --enable-debug=partial" exists. CC_FOR_BUILD is now needed for snowball as well as lemon. Fri Mar 30 19:53:05 BST 2007 Olly Betts <olly@survex.com> * common/utils.cc: Avoid GCC warning on format string "%I64d". Fri Mar 30 16:48:46 BST 2007 Richard Boulton <richard@lemurconsulting.com> * common/tcpserver.h,net/tcpserver.cc: Patch from Mark Hammond: implement for windows. Fri Mar 30 16:21:03 BST 2007 Richard Boulton <richard@lemurconsulting.com> * common/tcpclient.h,net/tcpclient.cc: Patch from Mark Hammond: implement for windows. Fri Mar 30 15:52:57 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/backendmanager.cc: Patch from Mark Hammond: implement launcher for tcpserver, so that remotetcp tests can run on windows. Fri Mar 30 15:42:49 BST 2007 Richard Boulton <richard@lemurconsulting.com> * common/safewindows.h: Include xapian/error.h so WinsockInitialiser can throw an exception. Fri Mar 30 14:59:40 BST 2007 Richard Boulton <richard@lemurconsulting.com> * More code from Mark Hammond's patch * common/remoteconnection.h: Add explicit destructor and (for windows compiles) a declare a helper method to calculate the number of milliseconds until a timeout. * net/remoteconnection.cc: Implement windows versions for functions which need it: constructor and destructor now do create and close the overlapped IO stuff, read_at_least() and send_message() have windows specific implementations. Fri Mar 30 13:05:27 BST 2007 Richard Boulton <richard@lemurconsulting.com> * More code adapted from Mark Hammond's patch * common/safewindows.h: When remote databases are compiled in, include the winsock stuff, and define some useful bits and pieces for using it. * common/remoteconnection.h: Add a member to RemoteConnection on windows to hold an overlapped IO structure. * net/progclient.cc: Don't include winsock stuff directly (it's now brought in by "safewindows.h"), and don't include <sys/wait.h> on windows. * net/remoteserver.cc: Ignore more SIGPIPE stuff on windows. Fri Mar 30 12:37:15 BST 2007 Richard Boulton <richard@lemurconsulting.com> * Patch from Mark Hammond * backends/dbfactory_remote.cc: Ignore code to handle SIGPIPE on windows, where it doesn't exist. * backends/remote/remote-database.cc: Raise UnimplementedError on windows if a progclient backend is requested. Fri Mar 30 12:11:50 BST 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/testsuite.cc: Fix srcdir guessing on windows - search for windows style directory separators, and remove .exe suffix from executable name. Adapted from Mark Hammond's patch. Thu Mar 29 18:32:18 BST 2007 Richard Boulton <richard@lemurconsulting.com> * Changes related to Mark Hammond's patch: * common/safeerrno.h: Define a new inline function "socket_errno()" which returns an error number relating to the last error caused by a socket function on platforms which make such a distinction, and the value of errno on other platforms. * net/tcpclient.cc,net/tcpserver.cc: Use the socket_errno() function instead of checking the value of errno. Also, check for WSAEWOULDBLOCK instead of EINPROGRESS on windows. Thu Mar 29 15:36:45 BST 2007 Richard Boulton <richard@lemurconsulting.com> * bin/xapian-tcpsrv.cc,tests/harness/testsuite.cc: First of many parts of a large patch from Mark Hammond working towards enabling remote databases on windows. When displaying errors which might be socket errors, display the error number as well as the output of strerror - on windows, strerror doesn't display useful information for socket errors. * AUTHORS: Add Mark Hammond Thu Mar 29 06:07:35 BST 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Unserialisation shouldn't reject a serialised query which ends after a termname. * tests/internaltest.cc: Add regression test. Wed Mar 28 17:17:42 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add feature tests for "prefer_mdy" and "epoch" arguments of DateValueRangeProcessor. Wed Mar 28 04:52:34 BST 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add more test cases for parsing of value ranges. Wed Mar 28 04:18:14 BST 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c,languages/compiler/header.h, languages/steminternal.cc,languages/steminternal.h: For among with functions (only used by finnish and lovins stemmers currently), change to generating an array of unsigned byte offsets into an array of function pointers rather than just an array of function pointers - this generally requires a lot less space. Wed Mar 28 03:28:37 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix parsing of `hello a..b'. * tests/queryparsertest.cc: Add regression test for above. Refactor qp_value_range1 to use a table of queries. Wed Mar 28 00:35:50 BST 2007 Olly Betts <olly@survex.com> * languages: Update svn:ignore for recent romanian changes. Wed Mar 28 00:08:11 BST 2007 Olly Betts <olly@survex.com> * api/valuerangeproc.cc: Need #include <stdio.h> for snprintf or sprintf. Tue Mar 27 23:53:18 BST 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix infinite loop in value range processing. Fix value range to work as a filter when used with a probabilistic query. * tests/queryparsertest.cc: Add regression tests for the above two issues. Refactor qp_value_range2 to use a table of queries. Tue Mar 27 22:09:57 BST 2007 Olly Betts <olly@survex.com> * api/valuerangeproc.cc: Assume that the start date is before the end date to help decide ambiguous cases. Tue Mar 27 21:49:15 BST 2007 Olly Betts <olly@survex.com> * api/valuerangeproc.cc: Refactor to simplify and reduce duplication. Tue Mar 27 21:41:38 BST 2007 Olly Betts <olly@survex.com> * configure.ac: datarootdir is new in 2.60 too, so use datadir when setting docdir for 2.59. Tue Mar 27 18:39:11 BST 2007 Olly Betts <olly@survex.com> * api/Makefile.mk,api/valuerangeproc.cc,include/xapian/queryparser.h: Move NumberValueRangeProcessor::operator()() out of the header as it's too complex for sane inlining. Implement handling of dmy and mdy dates and epochs for 2 digit years. * tests/queryparsertest.cc: Add tests. Tue Mar 27 11:57:53 BST 2007 Richard Boulton <richard@lemurconsulting.com> * languages/romanian.sbl: Add Martin Porter's new romanian stemming algorithm. * languages/romanian1.sbl,languages/romanian2.sbl: Remove old romanian stemming algorithms. * languages/Makefile.mk: Replace romanian1 and romanian2 by romanian. * languages/stem.cc: Use romanian stemmer when asked for "ro" or "romanian". No longer give a stemmer for "romanian1" or "romanian2". Tue Mar 27 04:50:36 BST 2007 Olly Betts <olly@survex.com> * languages/german2.sbl: Copy over "hop 3" in utf-8 case from snowball's latest version. Tue Mar 27 04:44:52 BST 2007 Olly Betts <olly@survex.com> * tests/stemtest.cc: Remove lower casing of dictionary words before stemming - instead we'll make sure the dictionary is already the right case. Tue Mar 27 04:28:10 BST 2007 Olly Betts <olly@survex.com> * languages/compiler/analyser.c,languages/compiler/generator.c, languages/compiler/header.h,languages/steminternal.cc, languages/steminternal.h: Add handling of among with functions. We use a "shim" functions for each method we want to call in this way. The shim function is passed the "this" pointer and calls a particular method on this. Split the functions into an optional separate array, since they're not used by most of the stemming algorithms. * languages/Makefile.mk,languages/stem.cc: Enable the finnish and lovins stemmers. Tue Mar 27 04:24:37 BST 2007 Olly Betts <olly@survex.com> * languages/steminternal.cc: Fix typo bug in slice_to - if we needed to increase the capacity of a snowball string variable, we trashed the variable with p (bug discovered while trying to enable the finnish stemmer). Tue Mar 27 02:55:51 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Implement handling of ISO format dates in DateValueRangeProcessor. * tests/queryparsertest.cc: Add feature tests. Tue Mar 27 01:47:36 BST 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Implement prefix and suffix handling for NumberValueRangeProcessor. * tests/queryparsertest.cc: Add feature tests. Mon Mar 26 22:26:48 BST 2007 Olly Betts <olly@survex.com> * queryparser/utf8itor.h: Add is_currency() predicate function. * queryparser/queryparser.lemony: Rejig parsing of RANGE_START to allow all characters which are likely to be required. * tests/queryparsertest.cc: Test RANGE_START can contain "/". * queryparser/queryparser.lemony: Set "syntax=yacc" for vim, which give fairly decent syntax highlighting. Mon Mar 26 15:43:50 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Use m4 comments not shell (dnl not #). Mon Mar 26 15:41:41 BST 2007 Olly Betts <olly@survex.com> * configure.ac: Add code to ensure that docdir is set for autoconf 2.59 (starting from 2.60, it is defined as standard). * docs/Makefile.am: Use docdir for installing docs. Sun Mar 25 15:41:16 BST 2007 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Prune unused docids off the start of each source database's range of docid. Sat Mar 24 23:07:42 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/base.h: Add a note explaining why it's OK that RefCntBase doesn't have a virtual destructor. Sat Mar 24 01:30:17 GMT 2007 Olly Betts <olly@survex.com> * HACKING,configure.ac: Bump the automake required version to 1.9.5. It was in Debian sarge, so any modern Linux distro should include packages, and we've actually been bootstrapping with 1.9.6 for ages. Thu Mar 22 00:53:35 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Disable probes for f77 more completely by preventing the probe code from even appearing in configure. Similarly eliminate the code for gcj and rc probes - in total these changes reduce the size of configure by 209KB (~25%). Fri Mar 09 14:28:01 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * Makefile.am: Add include/xapian/error.h to list of headers to install. Thu Mar 08 17:16:08 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * generate-exceptions.in,exception_data.pm: Split definitions of the exception hierarchy into a separate file, so it can be reused elsewhere. * Makefile.am: Add exception_data.pm as a dependency for generating error.h, and add it to EXTRA_DIST. Fri Mar 09 06:42:40 GMT 2007 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.cc,backends/flint/flint_cursor.h: We always ignore the return value from FlintCursor::get_key() because we only call it when it shouldn't fail, so restructure the code so it can't fail and change the return value to void. Fri Mar 09 06:33:09 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Suppress another aCC warning. Thu Mar 08 20:39:46 GMT 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Allow arbitrary characters in a range end (anything except whitespace and ')'). Thu Mar 08 19:46:46 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,languages/Makefile.mk: Depend on the autoconf substituted version of a script, not the template version. The only thing we substitute is @PERL@, but configure won't update a substituted file which hasn't changed so it's better to put the dependency on the script itself where it belongs. Thu Mar 08 17:09:21 GMT 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Keep any non-whitespace, non-term prefix for a term in case it turns out to be the start of a range, so "$50..100" works. Thu Mar 08 16:21:14 GMT 2007 Olly Betts <olly@survex.com> * HACKING: Add my "create docs/GNUmakefile" trick to avoid doxygen delays. Thu Mar 08 16:13:39 GMT 2007 Olly Betts <olly@survex.com> * HACKING: A few more coding guidelines. Thu Mar 08 14:26:21 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Adjust wording to indicate that the build system is indended not to break after SVN updates. Thu Mar 08 07:34:22 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,tests/queryparsertest.cc: Add simple implementations of DateValueRangeProcessor and NumberValueRangeProcessor and add qp_value_range2 to check that chaining ValueRangeProcessor subclasses works as intended. Thu Mar 08 06:06:15 GMT 2007 Olly Betts <olly@survex.com> * docs/overview.html: Fix links to error classes in generated API documentation. Thu Mar 08 05:37:13 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.cc, queryparser/queryparser.lemony,queryparser/queryparser_internal.h, tests/queryparsertest.cc: Add the ability to parse "value ranges" to the QueryParser (e.g. "10..50") which translate into a Query with operator OP_VALUE_RANGE. Thu Mar 08 04:27:54 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.cc: Remove QueryParser::set_stemming_options() which has been deprecated since 0.9.0. Thu Mar 08 03:37:07 GMT 2007 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc: Handle write failing to send a single byte down down the pipe to tell the parent we've got the lock (a potential bug noticed by _FORTIFY_SOURCE!) Handle read() failing with an unexpected error on the parent side by failing the lock attempt rather than risking an infinite loop. Thu Mar 08 02:52:26 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/types.h: Rewritten, with more accurate documentation comments. Add constant Xapian::BAD_VALUENO which is -1 cast to Xapian::valueno. * api/omenquire.cc,include/xapian/enquire.h,matcher/multimatch.cc: Use Xapian::BAD_VALUENO instead of Xapian::valueno(-1). Thu Mar 08 01:47:08 GMT 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Try putting apidoc/html in EXTRA_DIST instead of using dist-hook. Wed Mar 07 23:16:52 GMT 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Stop shipping docs/apidoc/latex/* in the xapian-core tarballs since it's just useless bloat. Removing it more than halves the size of the tarball (55% reduction!) Wed Mar 07 22:07:31 GMT 2007 Olly Betts <olly@survex.com> * tests/runtest.in: If we aren't using valgrind, turn on MALLOC_CHECK_ and MALLOC_PERTURB_ for glibc. Wed Mar 07 20:02:31 GMT 2007 Olly Betts <olly@survex.com> * INSTALL: CVS -> SVN. Wed Mar 07 20:00:37 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Define _FORTIFY_SOURCE in config.h if GCC is in use and it's not already set. That way the user can easily override. * INSTALL: Document this. Wed Mar 07 19:53:15 GMT 2007 Olly Betts <olly@survex.com> * INSTALL: Remove reference to muscat36 backend. Wed Mar 07 17:37:07 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Note GCC and glibc versions needed for _FORTIFY_SOURCE support. Wed Mar 07 16:29:02 GMT 2007 Olly Betts <olly@survex.com> * queryparser/lemon.c: Increase the length allowed for the filename in an error message, as it was being truncated which makes for a confusing error message and stops editors jumping to the line with the error in. Wed Mar 07 15:57:15 GMT 2007 Olly Betts <olly@survex.com> * common/Makefile.mk: Ship common/msvc_posix_wrapper.h and common/msvc_posix_wrapper.cc. Wed Mar 07 10:01:31 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Add -D_FORTIFY_SOURCE to AM_CXXFLAGS for GCC builds. According to glibc CVS support for this was added in October 2004, but it doesn't seem to be documented very well, other than in features.h. This adds some extra checking for array bounds, partially at compile time, but doesn't currently find any problems (or cause any noticeable slowdown). Wed Mar 07 04:49:01 GMT 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Include str_parameter in swap() and the copy constructor. * tests/api_anydb.cc: get_mset(0, 20) not get_mset(1, 20)! Also check that we didn't miss any documents which should match the filter, not just that we didn't get any documents which shouldn't. Wed Mar 07 03:07:49 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/unixcmds.h: Fixed reverse sense include guard test. * tests/api_db.cc: Still need #include "utils.h". * tests/harness/backendmanager.cc: Update for rmdir.h -> unixcmds.h and rmdir() -> rm_rf(). Wed Mar 07 02:22:30 GMT 2007 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/quartz_database.cc, backends/quartz/quartz_log.cc,common/utils.h,net/remoteconnection.cc, net/tcpclient.cc,tests/api_db.cc,tests/btreetest.cc,tests/harness/, tests/quartztest.cc: Rename rmdir() to rm_rf() since it does the same as "rm -rf" not "rmdir". The "touch()" function is only used in the testsuite so move it from common/utils.h into tests/harness/rmdir.cc and rename rmdir.cc to unixcmds.cc. common/utils.h no longer include safefcntl.h, so add explicit includes to the files which were previous relying on utils.h pulling it in. Wed Mar 07 02:21:18 GMT 2007 Olly Betts <olly@survex.com> * xapian.spec.in: Include ChangeLog.examples. Wed Mar 07 02:17:59 GMT 2007 Olly Betts <olly@survex.com> * docs/index.html,docs/internals.html: Create a separate index page for the "internal" documentation. Tue Mar 06 23:47:58 GMT 2007 Olly Betts <olly@survex.com> * common/utils.cc,common/utils.h: Remove map_string_to_value() which is no longer used. * common/utils.h: Remove prototype for rmdir() which I failed to remove with the earlier commit. Tue Mar 06 22:51:51 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/rmdir.cc: Fix rmdir() on Unix to not be O(n^2) in the worst case. Tue Mar 06 22:46:30 GMT 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Use the "inline test || echo" idiom to merge the VPATH and non-VPATH cases for distributing and installing the HTML apidocs. Tue Mar 06 22:13:03 GMT 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Fix Query::get_description() on an OP_VALUE_RANGE query. Tue Mar 06 09:03:33 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Note on running preautoreconf and autoreconf to keep SVN builds working. * PLATFORMS: Update with success reports for windows builds. Tue Mar 06 01:57:23 GMT 2007 Olly Betts <olly@survex.com> * tests/internaltest.cc: Comment out the "unsigned long long" testcase in tostring1 since the library doesn't currently have (or need) the utility function for that case. Tue Mar 06 01:49:55 GMT 2007 Olly Betts <olly@survex.com> * common/utils.cc,tests/btreetest.cc,tests/harness/, tests/quartztest.cc: rmdir() is only used in the test suite, and not in the library code, so move it into the testsuite. Also, bow out early if the pathname passed is empty, and on Unix protect against filenames which start with "-". Mon Mar 05 02:52:49 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Suppress more unhelpful aCC warnings. Mon Mar 05 02:15:29 GMT 2007 Olly Betts <olly@survex.com> * tests/internaltest.cc: Add testcase autoptr1 which tests that an autoptr correctly handles self-assignment. Mon Mar 05 02:05:53 GMT 2007 Olly Betts <olly@survex.com> * tests/internaltest.cc: Rename "omtostring1" to "tostring1" (we don't want to perpetuate references to "om"!) Add more test cases to this test. Mon Mar 05 01:18:17 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Update in line with 0.9.10. Sun Mar 04 23:58:29 GMT 2007 Olly Betts <olly@survex.com> * AUTHORS: Add reporters of bugs fixed in 0.9.10 to the "thanks" list. Sun Mar 04 00:44:17 GMT 2007 Olly Betts <olly@survex.com> * api/errorhandler.cc,include/xapian/errorhandler.h: Move virtual dtor for ErrorHandler out of the header. Same reasons as previous commit. Sun Mar 04 00:40:44 GMT 2007 Olly Betts <olly@survex.com> * api/omenquire.cc,include/xapian/enquire.h,matcher/Makefile.mk, matcher/weight.cc: Move virtual method definitions out of headers for MatchDecider, ExpandDecider, Weight, BoolWeight. This fixes warnings from aCC, and it's rare that the compiler would be able to inline a virtual method anyway. Sat Mar 03 21:40:43 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: For HP's aCC, we no longer need to suppress warning #336 (empty translation unit) since dummy.cc is history. But we do want to suppress a couple of frequently reported "remarks" which are unavoidable and don't indicate a problem. Experimentally, add "+wlint" for "link-like" warnings. Sat Mar 03 21:38:52 GMT 2007 Olly Betts <olly@survex.com> * include/xapian.h,include/xapian/version_h.cc: Move function prototypes into a non-generated header. * configure.ac,msvc/version.h.in: Generate version.h suitable for use with MSVC. * Makefile.am: Distribute msvc/version.h. Sat Mar 03 19:49:11 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Rework how we start and handle xapian-tcpsrv - it's now started by exec from a child process. This means that the mechanism to try higher port numbers works reliably, and we no longer leak file descriptors (which was causing tests to fail on some platforms). Sat Mar 03 19:46:03 GMT 2007 Olly Betts <olly@survex.com> * bin/xapian-tcpsrv.cc: Put "e.get_errno()" in a variable for clearer code. Sat Mar 03 16:01:17 GMT 2007 Olly Betts <olly@survex.com> * common/msvc_posix_wrapper.cc: Change `#include <fcntl.h>' to `#include "safefcntl.h"' (currently makes no difference on Windows, but it's more consistent to always use the safeXXX headers). * common/msvc_posix_wrapper.cc: Merge two ECHILD cases. * common/msvc_posix_wrapper.cc: Make sure dwCreationDisposition is initialised even if flags has a bogus value. * common/msvc_posix_wrapper.cc: Be consistent and always use O_CREAT, etc, instead of sometimes using the MS alternative forms _O_CREAT, etc. Sat Mar 03 03:25:05 GMT 2007 Olly Betts <olly@survex.com> * generate-exceptions.in: The GCC visibility docs aren't clear whether you need to make *thrown* exception types visible, or also *caught* exception types. But I've just been having odd problems catching Xapian::DocNotFoundError with "const Xapian::Error &", so let's mark the exception virtual baseclass Xapian::Error as visible too. Fri Mar 02 22:34:04 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk,languages/generate-allsnowballheaders.in: Convert unreadable make rule for generating allsnowballheaders.h to a perl script. * configure.ac: Substitute languages/allsnowballheaders. Fri Mar 02 14:31:11 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/internaltest.cc: Add a test for om_tostring(), mainly aimed at 64 bit types on windows platforms. Mingw builds report a warning in this code, which merits further investigation - see Bug #112 for details. Fri Mar 02 14:15:44 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/msvc_posix_wrapper.cc: Improve accuracy of emulation of posix open, and fix a bug which caused it to misinterpret the "flags" argument - files were always being opened for writing: O_RDONLY is 0, so the old test (flags & O_RDONLY) always failed. Fri Mar 02 12:44:33 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/msvc_posix_wrapper.cc: Fix whitespace issues. Fri Mar 02 12:18:42 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * HACKING: Add note about how to generate ChangeLog timestamps using the unix date command - and I've started generating them in the same format as Olly is. (I hope.) Fri Mar 2 11:49:11 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/msvc_posix_wrapper.h,common/safeerrno.h, common/msvc_posix_wrapper.cc,backends/quartz/btree.cc, backends/flint/flint_table.cc,backends/flint/flint_btreebase.cc: Move MSVC specific code for opening files into msvc_posix_wrapper.cc, and just leave a few conditional includes and function calls elsewhere. Thu Mar 01 22:51:42 GMT 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Add casts to U_isupper(), etc to suppress warnings from aCC. * queryparser/queryparser.lemony: Use U_isdigit() instead of C_isdigit(), etc - the truncation to a char can cause false positives. * tests/queryparsertest.cc: Add a (rather contrived) regression test for the above change. Thu Mar 01 21:49:37 GMT 2007 Olly Betts <olly@survex.com> * queryparser/utf8itor.h: Add cast to suppress warning from aCC. Thu Mar 1 15:27:09 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/utils.c,common/utils.h: For windows, add an om_tostring() function for 64 bit integers: time() and GetProcessId() return these, and without this, backends/quartz/quartz_log.cc reports an error. Thu Mar 1 13:48:46 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.lemony: Fix handling of hated wildcards which don't expand to any terms. Thu Mar 1 13:45:04 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/queryparsertest.cc: Fix last two test cases for hated wildcards to check for exactly what should be returned. Thu Mar 1 12:02:04 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * common/safeerrno.h: Some of Microsoft's C++ compiler versions earlier than 2005 do not have _set_errno, so #define it in this case. Thu Mar 01 04:28:23 GMT 2007 Olly Betts <olly@survex.com> * backends/quartz/btree.h: Correct comment ("read" -> "write"). Thu Mar 01 04:21:52 GMT 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.h: Correct comment ("read" -> "write"). Thu Mar 01 01:30:02 GMT 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add some test cases for hated wildcards which don't expand to any terms. Thu Mar 01 01:21:42 GMT 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Complete truncated comment. Prefer preincrement to postincrement. Wed Feb 28 20:13:35 GMT 2007 Olly Betts <olly@survex.com> * api/omdocument.cc: Really tweak OmDocumentTerm::add_position() so that adding position 0 to an empty termlist takes the shortcut. Wed Feb 28 15:04:25 GMT 2007 Olly Betts <olly@survex.com> * common/safeerrno.h: Correct <safeerrno.h> to "safeerrno.h" in message in #error directive. Wed Feb 28 14:59:29 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/query.h: Removed documentation comment for Xapian::Query::Internal which is no longer relevant (Query objects are now immutable once constructed, so parameters can no longer be set separately). Wed Feb 28 02:42:31 GMT 2007 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Add support for serialising OP_VALUE_RANGE queries. Add support for OP_VALUE_RANGE queries to Query::get_description(). If an OP_VALUE_RANGE query has a start bound strictly greater than its end bound, simplify it by knowing that it can't match anything. * common/remoteprotocol.h,docs/remote_protocol.html: Bump remote protocol version to 24. Wed Feb 28 01:15:00 GMT 2007 Olly Betts <olly@survex.com> * api/omquery.cc,api/omqueryinternal.cc,include/xapian/query.h, matcher/: Add new Query operator OP_VALUE_RANGE which allows a query to be filtered by checking if a value lies between two limits. * tests/api_anydb.cc: Add feature test valuerange1. Wed Feb 28 01:07:53 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Only hook the testsuite into valgrind if env var XAPIAN_TESTSUITE_VALGRIND is set, to make it simple to run a test program under valgrind normally. * tests/runtest.in: Set XAPIAN_TESTSUITE_VALGRIND if automatically running a test program under valgrind. Tue Feb 27 23:51:34 GMT 2007 Olly Betts <olly@survex.com> * tests/Makefile.am: Turn subdir-objects back on here so that the harness objects end up in the harness subdirectory. Tue Feb 27 22:45:04 GMT 2007 Olly Betts <olly@survex.com> * examples/delve.cc: Rename "-k" to "-V" since "keys" were renamed to "values" long ago. Keep "-k" as an alias for now, but don't advertise it. Add handling so "-V3" shows value #3 for every document in the database. Tue Feb 27 21:35:35 GMT 2007 Olly Betts <olly@survex.com> * tests/Makefile.am,tests/remotetest.cc: Get rid of remotetest since it no longer does anything. Tue Feb 27 21:28:28 GMT 2007 Olly Betts <olly@survex.com> * tests/api_db.cc,tests/remotetest.cc: Move netstats1 from remotetest to apitest. It will now run for "remotetcp" as well as "remote". Tue Feb 27 21:27:17 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc,tests/harness/backendmanager.h: Inline trivial getter and setter methods in the header. Make getter method const. Tue Feb 27 20:22:58 GMT 2007 Olly Betts <olly@survex.com> * examples/delve.cc: Change to not stem terms by default. Add "-s/--stemmer" option to allow a stemmer to be specified. Tue Feb 27 17:50:13 GMT 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Test revision_supplied rather than revision_ so that behaviour matches comments. Tue Feb 27 16:58:56 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_wrdb.cc: Enhance crashrecovery1 to check that the readers reference a database with the expected number of documents. This checks that the base files which were removed were the correct ones to leave the current database valid. Tue Feb 27 16:54:31 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * backends/quartz/btree.cc,backends/quartz/btree.h, backends/quartz/quartz_database.cc: Apply fix for opening databases while under heavy modification to quartz, using Olly's attachment to bug #108: http://www.xapian.org/cgi-bin/bugzilla/attachment.cgi?id=43 Verified with loadtest that this resolves the issue. Tue Feb 27 13:12:31 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.h: Add get_dbtype() method to allow the current backend type to be read. * tests/apitest.cc,tests/apitest.h: Add get_dbtype() function to allow BackendManager::get_dbtype() to be called by tests. * tests/api_wrdb.cc: Add new test crashrecovery1 to verify that the backend can recover from the latest revision missing some of the base files. Tue Feb 27 12:45:00 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Fix how we start xapian-progsrv when running under valgrind. Tue Feb 27 11:34:22 GMT 2007 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Remove the old throw for DatabaseOpeningError so the new throw for DatabaseModifiedError will actually be used. Tue Feb 27 09:22:52 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * flint_table.cc,flint_table.h,flint_database.cc: If a database is being frequently modified (ie, transactions committed extremely frequently), flint can fail open all the tables at a consistent revision on the first attempt. The code to handle this in flint_database.cc was not firing because flint_table.cc was reporting all failures to open a table for reading by throwing an exception. This commit changes flint_table.cc to report failure to open at a specific revision by returning false. In addition, there was a bug in flint_database.cc which caused the attempts to re-try opening a database in this situation to fail. This is now fixed. Finally, flint_database.cc now throws a DatabaseModified error if 100 attempts to reopen the database fail because of continual modifications - previously, it would have thrown a DatabaseOpening error. There is no easy regression test - but my "loadtest.cc" program attached to bug #108 shows the problem (if you comment out the code which ignores DatabaseOpeningErrors). Tue Feb 27 01:25:00 GMT 2007 Olly Betts <olly@survex.com> * docs/doxygen_api_conf.in,docs/doxygen_full_conf.in: Need to define DOXYGEN so that the previous change to the generated errordispatch.h works. Remove obsolete macros from EXPAND_AS_DEFINED and PREDEFINED. * docs/doxygen_full_conf.in: Increase MAX_DOT_GRAPH_WIDTH from 800 to 1024 to match doxygen_api_conf.in. Tue Feb 27 00:41:47 GMT 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix warning with SGI's CC. Tue Feb 27 00:35:06 GMT 2007 Olly Betts <olly@survex.com> * generate-exceptions.in: Wrap errordispatch.h in "#ifndef DOXYGEN" and "#endif" because doxygen gets confused by a header full of code. Mon Feb 26 21:53:41 GMT 2007 Olly Betts <olly@survex.com> * tests/findheaders.pl: Remove long unused perl script. Mon Feb 26 19:23:50 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/api_wrdb.c: test_emptyterm2 used to delete a writable database whilst the database was open. Unfortunately, this doesn't work on windows (and it doesn't seem to be easy to make it work), so this patch changes the test to close each database before opening the next writable database at the same path. Fortunately, this seems to be the only test which behaves like this. Also, added a note to the win32 README file to document this restriction. Mon Feb 26 19:19:34 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * docs/Makefile.am: touch apidoc/html/index.html after apidoc.pdf is created so that make doesn't attempt to regenerate it due to the dependency of apidoc/html/index.html on apidoc.pdf. Similarly for sourcedoc/html/index.html. Mon Feb 26 18:36:22 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: Indent with tabs not spaces. * tests/harness/index_utils.cc: Add missing space after "if"; prefer "X.empty()" to "X.size() == 0". Mon Feb 26 16:10:31 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * tests/harness/index_utils.cc,tests/harness/backendmanager.cc: Fix for windows: windows file handling seems to have a bug causing end of file conditions in ifstream to be missed. As a result, an empty string was being returned from the get_paragraph function in the test harness, resulting in an out-of-bounds error when generating a value from the paragraph. Add a check that the paragraph returned isn't empty, and return an empty document if it is. Then, check for empty documents (ie, contain some terms) before adding them to the test database. Mon Feb 26 16:05:07 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * backends/quartz/btree_base.cc,backends/flint/flint_btreebase.cc: Fix for windows: string append method with a pointer and length causes a segfault-type error on windows if the pointer is NULL, even if the length is also 0. Check for this case when appending the bitmap to the base block, since a NULL pointer is passed for an empty bitmap when a new database is being created. Also, initialise the bit_map0 and bitmap members to 0 in the FlintTable_base constructor which takes a string and a char argument - they had been omitted from this and were only being initialised by the constructor which takes no argument. This didn't seem to be having any negative effect, but doing it makes the code neater. Sun Feb 25 23:21:11 GMT 2007 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc: Create the RefCntPtr to this as an explicit variable to make the code more readable. Sun Feb 25 19:20:33 GMT 2007 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix inconsistent ordering between pages with set_sort_by_value_then_relevance (fixes bug#110). * tests/api_wrdb.cc: Enabled test consistency2. Sun Feb 25 17:28:09 GMT 2007 Olly Betts <olly@survex.com> * matcher/: Also apply the "decreasing weights with remote database" optimisations which we use in the sort_by_relevance case in the sort_by_relevance_then_value case. Sat Feb 24 19:35:41 GMT 2007 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc: If replace_document is used to set the docid of a newly added document which has previously existed, then we need to mark that document as valid (caught by existing test replacedoc4 in a debug build). Sat Feb 24 17:11:22 GMT 2007 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Add regression test consistency2 for bug#110 (currently skipped since the bug isn't fixed yet). Thu Feb 22 15:18:53 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Add proper detection for SGI's C++ (check stderr output of "CC -v"). Automatically pass -ptused in CXXFLAGS for xapian-core and any applications using xapian-config --cxxflags since it seems to be required to avoid template linking errors. Thu Feb 22 09:04:30 GMT 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Ship generated RSTHTML files. Wed Feb 21 19:01:06 GMT 2007 Olly Betts <olly@survex.com> * languages/turkish.sbl: Make some simplifications to the code of the turkish stemmer. Wed Feb 21 19:00:11 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk,languages/stem.cc,languages/turkish.sbl: Add turkish stemmer. Wed Feb 21 18:55:03 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Number m_test and c_test variables to avoid variable shadowing warnings. Wed Feb 21 18:07:50 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Remove duplicate copy of function resulting from ham-fisted reverting and applying of patches. Wed Feb 21 18:03:12 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c,languages/compiler/header.h: Number mlimit variables to avoid variable shadowing warnings. * languages/compiler/generator.c: Use "if (foo() == -1) return -1;" which avoids a temporary variable (improving readability of generated code) and producing very slightly faster code. * languages/compiler/header.h: Use "#ifdef DISABLE_JAVA" around Java generator specific struct members and prototypes. Wed Feb 21 17:47:13 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Add "~C" comments to generated source for more operations. Tue Feb 20 20:24:30 GMT 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc,backends/quartz/btree.cc: Ensure both_bases is set to false if we don't have both bases when opening a table using an existing object. Tue Feb 20 15:57:05 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Fix to handle generating dependencies on generated objects, at least for the "checked out tree from SVN" case. Fixing this fully is trickier, but currently this is stopped snapshots from bootstrapping. Tue Feb 20 14:28:02 GMT 2007 Charlie Hull * backends/flint/flint_btreebase.cc,backends/flint/flint_table.cc, backends/quartz/btree.cc: Use MS Windows API calls to delete files and open files we might want to delete while they are still open (i.e. the flint and quartz btree base files). This fixes a problem when a writer can't discard an old revision at the exact moment a reader is opening it (bug #108). Tue Feb 20 14:13:43 GMT 2007 Richard Boulton <richard@tartarus.org> * .: Add generate-exceptions to svn:ignore property. Tue Feb 20 13:36:27 GMT 2007 Olly Betts <olly@survex.com> * generate-exceptions.in: Add missing '}' to close 'namespace Xapian' in generated header 'xapian/error.h'. Tue Feb 20 13:26:19 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am: Put generated exception headers in BUILT_SOURCES. Tue Feb 20 12:10:18 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,generate-exceptions.in,include/Makefile.mk, include/xapian/error.h,include/xapian/errortypes.h,net/serialise.cc: Replace macro gymnastics with errortypes.h by a perl script which generates the headers we require. It's more flexible, and easier to get doxygen to generate documentation from. * queryparser/queryparser.cc,tests/queryparsertest.cc: Throw new Error subclass QueryParserError instead of throwing const char * (fixes bug#101). Tue Feb 20 10:37:54 GMT 2007 Olly Betts <olly@survex.com> * languages/steminternal.cc: p is never NULL once the class is successfully constructed, so just use Assert() to confirm this. Tue Feb 20 10:29:25 GMT 2007 Olly Betts <olly@survex.com> * queryparser/lemon.c: Fix compiler warnings. Mon Feb 19 18:50:12 GMT 2007 Olly Betts <olly@survex.com> * common/serialise-double.cc,net/remoteconnection.cc: Throw NetworkError rather than InternalError for invalid data received over the remote protocol. Mon Feb 19 18:12:21 GMT 2007 Olly Betts <olly@survex.com> * tests/internaltest.cc: Just disable serialisequery1 if the remote backend is disabled, like we do for serialiselength1 and serialisedoc1 rather than assuming InternalError is due to the code being disabled. Mon Feb 19 09:48:02 GMT 2007 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/XapianAdminNotes.txt,docs/admin_notes.rst: Generated file "XapianAdminNotes.html" doesn't fit the naming conventions of other files at all, so renamed to "admin_notes.html". Rename the source to ".rst" so we can use an implicit rule for the conversion, in anticipation of other documentation in ReST format. Mon Feb 19 09:10:48 GMT 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Fix typos which stopped parallel make from working correctly in "docs" subdirectory. Sun Feb 18 22:36:46 GMT 2007 Olly Betts <olly@survex.com> * common/Makefile.mk,common/omstringstream.h,common/output.h, include/Makefile.mk,include/xapian/output.h: Move "<xapian/output.h>" to being an internal header in "common/" since it hasn't been included by "<xapian.h>" since 0.7.0. Sun Feb 18 22:18:33 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/dbfactory.h: Add documentation comment for the "prog" form of Remote::open_writable(). Sun Feb 18 19:31:20 GMT 2007 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add test for "associative NEAR". Sun Feb 18 18:27:40 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc,tests/harness/backendmanager.h: Merge BackendManager::change_names_to_paths() into index_files_to_database() and make the latter a method of class BackendManager. * tests/harness/backendmanager.cc: Remove mention of muscat36 backends from error message listing valid backends. * tests/harness/backendmanager.h: Comment out prototypes for currently disabled methods getdb_inmemoryerr, getwritedb_inmemory, etc. Sat Feb 17 15:38:05 GMT 2007 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: new throws if allocation fails so we don't need to check the return value. Sat Feb 17 02:08:22 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_table.cc: Fix memory double-frees if FlintTable::close() is called twice on a table (without an open() being called in between). This was happening after a transaction failed when the table was next opened. * backends/quartz/btree.cc: Fix double-frees, similarly. Fri Feb 16 20:11:13 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Add missing "~Z" for compatibility with C code generation (which we don't use for Xapian). Fri Feb 16 19:51:26 GMT 2007 Olly Betts <olly@survex.com> * queryparser/tclUniData.h,queryparser/utf8itor.h: The tcl unicode routines only have tables for characters in the BMP. For other characters, assume they're word characters, but can't be forced to lowercase. Fri Feb 16 19:26:00 GMT 2007 Olly Betts <olly@survex.com> * queryparser/utf8itor.cc: Fix bug in decoding of 4 byte utf-8 sequences - the returned value was 0x400000 too large! Thu Feb 15 19:37:26 GMT 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Keep embedded apostrophe's in terms rather than generating a phrase search for them. Thu Feb 15 18:42:49 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c,languages/steminternal.cc, languages/steminternal.h: Tweak the "grouping" functions to allow skipping past multiple occurrences and use this to implement "goto" or "gopast" followed by a grouping or "non" grouping more efficiently. Thu Feb 15 17:57:43 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c,languages/steminternal.cc, languages/steminternal.h: Inline lose_s(). Since lose_s() checks if p is NULL, remove checks prior to calling lose_s(). * languages/steminternal.h: Cast from symbol * to int * via void * to try to suppress over-eager warnings about alignment. Thu Feb 15 12:43:04 GMT 2007 Olly Betts <olly@survex.com> * languages/stem.cc: Add support for two letter ISO 639 codes. Thu Feb 15 10:40:40 GMT 2007 Olly Betts <olly@survex.com> * docs/Makefile.am: Don't try to ship mkdoc.pl. Thu Feb 15 09:58:48 GMT 2007 Olly Betts <olly@survex.com> * docs/gen_codestructure_doc.in: Fix to work in a VPATH build. Thu Feb 15 09:54:55 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Fix Stem::get_description() to return the right string. Thu Feb 15 09:44:54 GMT 2007 Olly Betts <olly@survex.com> * docs/gen_codestructure_doc.in: Add links to viewvcs on svn.xapian.org. Thu Feb 15 09:26:07 GMT 2007 Olly Betts <olly@survex.com> * configure.ac,docs/Makefile.am,docs/gen_codestructure_doc.in, docs/mkdoc.pl,preautoreconf: Rework how code_structure.html is generated. Nicer output, and the html document is now regenerated if any of the dir_contents files change. Thu Feb 15 07:43:39 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,bin/Makefile.mk,configure.ac,examples/Makefile.mk, makemanpage.in: Add "makemanpage" script which takes care of all the magic needed to generate a man page using help2man. This script is written in perl, but so is help2man, and they're both only used in maintainer-mode. Thu Feb 15 06:33:41 GMT 2007 Olly Betts <olly@survex.com> * bin/Makefile.mk: Fix typos in man page generation rules. Thu Feb 15 06:30:54 GMT 2007 Olly Betts <olly@survex.com> * examples/Makefile.mk: Fix typos in previous commit. Thu Feb 15 06:03:35 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,examples/Makefile,examples/Makefile.am, examples/Makefile.mk: Convert examples subdirectory to use non-recursive make. Fixes wishlist bug#97. Thu Feb 15 05:48:38 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,bin/Makefile,bin/Makefile.am,bin/Makefile.mk, configure.ac: Convert bin subdirectory to use non-recursive make. Wed Feb 14 19:08:49 GMT 2007 Olly Betts <olly@survex.com> * tests/Makefile.am: There's no longer a need to remove the temporary directories for muscat36 backend tests, since we've removed the muscat36 backend code. Wed Feb 14 18:58:31 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk: Move adjusting of INCLUDES to the top of the file. * queryparser/Makefile.mk: Adjust INCLUDES for a VPATH build. Wed Feb 14 18:38:16 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk: Need -Ilanguages for VPATH builds from SVN. Wed Feb 14 18:37:54 GMT 2007 Olly Betts <olly@survex.com> * languages/steminternal.cc,languages/steminternal.h: Add SET_CAPACITY() macro. Wed Feb 14 17:39:50 GMT 2007 Olly Betts <olly@survex.com> * languages/steminternal.h: Don't cast away const. Wed Feb 14 17:14:14 GMT 2007 Olly Betts <olly@survex.com> * include/Makefile: Add stub makefile. Wed Feb 14 17:12:41 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: ac_abs_srcdir, etc aren't available in configure itself, so we have to do the work ourselves. Wed Feb 14 12:54:26 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Fix typo in last commit. Wed Feb 14 12:19:18 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,languages/Makefile.mk: Define automake conditional VPATH_BUILD so we can only add extra entries to INCLUDES in VPATH builds. Wed Feb 14 11:59:02 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk: Ick, we need $(top_srcdir)/languages in INCLUDES as the generated sources need to find non-generated headers. Wed Feb 14 11:46:43 GMT 2007 Olly Betts <olly@survex.com> * backends/database.cc,backends/remote/remote-database.cc, common/database.h,common/remote-database.h,common/remoteprotocol.h, docs/remote_protocol.html,net/remoteserver.cc: Support get_lastdocid() on remote databases. Wed Feb 14 11:20:33 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,include/Makefile.am,include/Makefile.mk: Convert include subdirectory to use non-recursive make. Wed Feb 14 11:05:00 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,bin/Makefile.am,configure.ac,examples/Makefile.am, getopt/Makefile,getopt/Makefile.am,getopt/Makefile.mk, tests/Makefile.am: Convert getopt subdirectory to use non-recursive make. Wed Feb 14 10:56:15 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/Makefile.am: Remove as no longer used. Wed Feb 14 10:38:27 GMT 2007 Olly Betts <olly@survex.com> * backends/quartz/Makefile.mk,bin/Makefile.am,tests/Makefile.am: Move backends/quartz/libbtreecheck.la to libquartzcheck.la. * getopt/Makefile.am: Distribute getopt/dir_contents. Wed Feb 14 10:20:14 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Handle variables set with "+=" and ":=". Wed Feb 14 10:16:29 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/database.h: Add documentation comments for DB_* constants. Wed Feb 14 10:11:37 GMT 2007 Olly Betts <olly@survex.com> * examples/.cvsignore: Remove lingering relic of CVS days. Wed Feb 14 10:07:14 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,net/Makefile,net/Makefile.am, net/Makefile.mk: Convert net subdirectory to use non-recursive make. Now libxapian.la is built completely non-recursively. * docs/Makefile.am,docs/mkdoc.pl: Fix mkdoc.pl to handle directories which use non-recursive make. Wed Feb 14 09:10:27 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,backends/Makefile,backends/Makefile.am, backends/Makefile.mk,backends/flint/Makefile, backends/flint/Makefile.am,backends/flint/Makefile.mk, backends/inmemory/Makefile,backends/inmemory/Makefile.am, backends/inmemory/Makefile.mk,backends/multi/Makefile, backends/multi/Makefile.am,backends/multi/Makefile.mk, backends/quartz/Makefile,backends/quartz/Makefile.am, backends/quartz/Makefile.mk,backends/remote/Makefile, backends/remote/Makefile.am,backends/remote/Makefile.mk, configure.ac: Convert backends subdirectory (and subdirectories thereof) to non-recursive make. Wed Feb 14 08:49:02 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Fix handling of backslash at the end of a line in included files. * preautoreconf: Our handling of which directory a nested include is sought from is consistent with automake, so remove the FIXME comment about it. Wed Feb 14 08:07:07 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,matcher/Makefile,matcher/Makefile.am, matcher/Makefile.mk: Convert matcher subdirectory to non-recursive make. Wed Feb 14 07:50:19 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,api/Makefile,api/Makefile.am,api/Makefile.mk, configure.ac: Convert api subdirectory to non-recursive make. Wed Feb 14 07:48:45 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: No need to `chdir' to the current directory. * preautoreconf: Don't bother looking at the file we generate when considering included files. * preautoreconf: Add included files to the dependency list. Wed Feb 14 07:38:19 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Generate dummy rules for the files which preautoreconf depends on, so that "make" still works if one of them is removed (by build system restructuring). Wed Feb 14 07:24:07 GMT 2007 Olly Betts <olly@survex.com> * common,languages,tests/harness: svn:ignore: We do need to ignore .dirstamp files after all. Wed Feb 14 07:22:29 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,common/Makefile,common/Makefile.am,common/Makefile.mk, configure.ac: Convert common subdirectory to non-recursive make. Wed Feb 14 07:03:27 GMT 2007 Olly Betts <olly@survex.com> * .: svn:ignore: Ignore all vim swapfiles. No need to ignore .deps now that dummy.cc is history. Wed Feb 14 06:34:44 GMT 2007 Olly Betts <olly@survex.com> * tests/stemtest.cc: Force ASCII capitals in sample vocabulary to lower case before stemming. Wed Feb 14 06:14:42 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,queryparser/Makefile, queryparser/Makefile.am,queryparser/Makefile.mk: Convert queryparser subdirectory to non-recursive make. * Makefile.am: Enable automake option "subdir-objects" so that built objects for a non-recursively made directory go into that directory. Wed Feb 14 06:12:29 GMT 2007 Olly Betts <olly@survex.com> * tests/harness: svn:ignore: remove .dirstamp and depcomp. Wed Feb 14 06:10:44 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk: In maintainer-mode, remove snowball and generated allsnowballheaders.h on "make clean". When not in maintainer-mode, remove them on "make maintainer-clean". * languages/: svn:ignore all vim swap files. Wed Feb 14 05:54:32 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk: Distribute the stub languages/Makefile. * tests/harness/Makefile.mk: No need to explicitly distribute harness/Makefile.mk. Wed Feb 14 05:52:08 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk: Distribute languages/dir_contents. Wed Feb 14 05:44:10 GMT 2007 Olly Betts <olly@survex.com> * api/Makefile.am,api/vectortermlist.h,common/Makefile.am, queryparser/Makefile.am: We use vectortermlist.h from the queryparser directory, so move it from api to common. Wed Feb 14 05:27:13 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk,languages/stem.cc: Disable finnish and lovins temporarily, until I sort out merging the support for among with functions properly. Wed Feb 14 05:24:36 GMT 2007 Olly Betts <olly@survex.com> * languages/steminternal.cc,languages/steminternal.h: Temporarily back out the changes to handling of among with functions. Tue Feb 13 19:11:57 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: If a test throws an unknwon exception, say so in the test failure message. If it throws std::string, report the first 40 character (or first line) of the string. Tue Feb 13 17:48:11 GMT 2007 Olly Betts <olly@survex.com> * languages/steminternal.cc: Fix decoding of 3 byte utf-8 sequences. Tue Feb 13 17:47:22 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk: Fix dependency of generated languages/allsnowballheaders.h to be on languages/Makefile.mk. Tue Feb 13 17:46:42 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c: Fix missing ";" in generated code. Tue Feb 13 15:40:15 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk,languages/compiler/driver.c, languages/compiler/generator.c,languages/compiler/header.h: Changes towards getting C++ generation working. Tue Feb 13 15:39:23 GMT 2007 Olly Betts <olly@survex.com> * languages/steminternal.cc,languages/steminternal.h: Revert to match currently generated code. Tue Feb 13 14:09:31 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/: Sync with current snowball SVN + submitted patches. Tue Feb 13 10:53:34 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/backendmanager.cc: If we fail to start xapian-tcpsrv, report why (give errno or the error output). Tue Feb 13 10:28:43 GMT 2007 Olly Betts <olly@survex.com> * tests/remotetest.cc: Prune test cases which are redundant now that apitest tests remote and remotetcp: apitest's simplequery1 performs a superset of what netmatch1 and tcpmatch1 do; apitest's multidb1 performs a superset of what netmatch2 does; and the coverage given by our netexpand1 is provided by various apitest tests. Mon Feb 12 06:36:58 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/generator.c: In backward mode, among can't match if there are fewer characters before the current position than the minimum length string in the among. Mon Feb 12 05:56:31 GMT 2007 Olly Betts <olly@survex.com> * configure.ac: Make sure that the languages subdirectory exists in case this is a fresh SVN checkout and srcdir != builddir. Mon Feb 12 05:46:49 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile.mk,languages/compiler/driver.c: -DDISABLE_JAVA to not compile in Java support (inline with Richard's patch to the upstream snowball sources). * languages/Makefile.mk: Remove inaccurate comment about processing with automake. Fix -n option we pass (remove directory name). Use "" instead of <> in generated file allsnowballheaders.h. Remove snowball compiler and generated files on "make maintainer-clean". Mon Feb 12 01:08:50 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * languages/compiler/header.h: Remove commented out header for sort() prototype, to match snowball sources. Sun Feb 11 19:28:50 GMT 2007 Olly Betts <olly@survex.com> * languages/api.h,languages/header.h: Remove headers which are no longer used. Sun Feb 11 19:26:50 GMT 2007 Olly Betts <olly@survex.com> * languages/api.cc,languages/utilities.cc: Remove sources which are no longer used. Sun Feb 11 16:22:16 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/driver.c: Use #ifdef instead of #if. Sun Feb 11 16:20:44 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/driver.c: Disable java support, since we don't use it and it means one less source file. Sun Feb 11 16:08:49 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/analyser.c,languages/compiler/header.h: Use qsort instead the snowball's own sort implementation to reduce the number of source files we need for the snowball compiler. Sun Feb 11 16:04:19 GMT 2007 Olly Betts <olly@survex.com> * languages/Makefile: Add stub Makefile to allow "make", "make check", and "make clean" from the languages subdirectory. Sun Feb 11 16:00:51 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/analyser.c: Cast the arguments, not the function pointer as ISO C allows `void*' and `struct amongvec*' to have different representations. Sun Feb 11 15:43:49 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/space.c: Fix a typo of a function name in a comment. Sun Feb 11 15:23:51 GMT 2007 Olly Betts <olly@survex.com> * languages/compiler/header.h,languages/compiler/tokeniser.c: Give syswords and syswords2 .h extensions for clarity. Sun Feb 11 15:17:28 GMT 2007 Olly Betts <olly@survex.com> * Makefile.am,api/Makefile.am,api/omstem.cc,configure.ac,dummy.cc, languages/Makefile.am,languages/Makefile.mk,languages/compiler/, languages/: Update the stemmers to use utf-8 and the latest snowball version (based on r421 from snowball SVN). Further patches are required to generate C++ classes from snowball, I'm checking in the vanilla compiler first to help us track the patches. Sun Feb 11 14:16:35 GMT 2007 Olly Betts <olly@survex.com> * api/omstem.cc: Add definition for Xapian::Stem::stem_word(). Sun Feb 11 13:15:11 GMT 2007 Olly Betts <olly@survex.com> * tests/api_anydb.cc,tests/api_db.cc,tests/api_posdb.cc: Prefer Xapian::Stem::operator() to Xapian::Stem::stem_word(). Sat Feb 10 19:10:31 GMT 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Prefer Xapian::Stem::operator() to Xapian::Stem::stem_word(). Sat Feb 10 18:34:41 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/stem.h: Make Xapian::Stem::Internal as @internal for doxygen. Mark Xapian::Stem::stem_word() as deprecated. Sat Feb 10 18:34:05 GMT 2007 Olly Betts <olly@survex.com> * HACKING: Add brief note about coding style. Sat Feb 10 18:24:28 GMT 2007 Olly Betts <olly@survex.com> * preautoreconf: Add licence and copyright notice. Add support for automake include directives. Warn about unknown variables. Sat Feb 10 18:19:26 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Conditionalise all the valgrind specific code on RUNNING_ON_VALGRIND to minimise the overhead if valgrind is installed but we aren't running tests under it. Sat Feb 10 17:42:05 GMT 2007 Olly Betts <olly@survex.com> * tests/harness/index_utils.cc,tests/stemtest.cc: Prefer Xapian::Stem::operator() to Xapian::Stem::stem_word(). Sat Feb 10 17:27:27 GMT 2007 Olly Betts <olly@survex.com> * examples/delve.cc,examples/simpleindex.cc,examples/simplesearch.cc: Prefer Xapian::Stem::operator() to Xapian::Stem::stem_word(). Wed Feb 07 04:23:09 GMT 2007 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,queryparser/queryparser.lt: Use new/delete instead of malloc/free to allocate struct yyParser and pass it around as "yyParser *" instead of "void *". Wed Feb 07 03:44:16 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.cc, queryparser/queryparser.lemony,queryparser/queryparser_internal.h, tests/queryparsertest.cc: Allow the default prefix to be specified so you can use QueryParser to parse a text entry box for "author" etc. * configure.ac: Note API change for LIBRARY_VERSION_INFO updating. Wed Feb 07 01:14:37 GMT 2007 Olly Betts <olly@survex.com> * common/safeunistd.h: Add '#include <sys/types.h>' so that we've seen a typedef for off_t before we hide it behind a #define to avoid a compilation error if <sys/types.h> or <wchar.h> is included after we are. Tue Feb 06 06:26:24 GMT 2007 Olly Betts <olly@survex.com> * INSTALL: Mention MSVC makefiles. Note that we still test build with GCC 2.95.4 even though we recommend using a newer version. Tue Feb 06 06:13:48 GMT 2007 Olly Betts <olly@survex.com> * AUTHORS,HACKING,backends/flint/,backends/quartz/btree.cc, backends/quartz/quartz_database.cc,bin/quartzcompact.cc, bin/xapian-compact.cc,common/,tests/btreetest.cc, tests/harness/backendmanager.cc,tests/harness/testsuite.cc, tests/quartztest.cc: Support large files in stat() and fstat() when building with MSVC. Factor MSVC specific header magic into new replacement headers "safeunistd.h" and "safesysstat.h". * AUTHORS: Thanks to Charlie Hull for the original patch to support stat() on large files with MSVC (and past MSVC related stuff!) Sat Jan 13 02:06:15 GMT 2007 Olly Betts <olly@survex.com> * include/xapian/database.h: Add pointer from WritableDatabase::add_document() to replace_document() in case the user wants to specify the docid themselves. Also improve the wording of the warning about changes not being committed to disk right away. Tue Jan 2 15:49:43 GMT 2007 Richard Boulton <richard@lemurconsulting.com> * backends/quartz/btree.cc,backends/flint/flint_io.h: Patches from Charlie Hull to allow 2GB+ index files work when compiled using Visual C++. Sat Dec 30 13:48:54 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.lemony,include/xapian/queryparser.h: Add support for partially entered queries, to better support incremental searching. Also fix a typo in a comment. * docs/queryparser.html: Document the new partial query option, and what it's useful for. * tests/queryparsertest.cc: Test partial queries. Wed Dec 20 16:42:24 GMT 2006 Olly Betts <olly@survex.com> * PLATFORMS: Updated with success reports from debian buildds. Wed Dec 20 00:53:19 GMT 2006 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Fix "unused parameter" warning when building with the remote backend disabled. Also update a comment - the query serialisation isn't really one line of text since we now store term names with the length prefixed, and they can contain arbitrary characters including linefeeds. Tue Dec 19 20:40:27 GMT 2006 Olly Betts <olly@survex.com> * tests/Makefile.am: Need libgetopt.la for test programs which use the test harness for platforms where GNU getopt isn't in the standard C library. Tue Dec 19 20:34:16 GMT 2006 Olly Betts <olly@survex.com> * queryparser/utf8itor.h: Need <string.h> for strlen. Tue Dec 19 16:35:58 GMT 2006 Olly Betts <olly@survex.com> * common/omstringstream.h: Fix "warning: comparison between signed and unsigned". Tue Dec 19 15:33:19 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Turn the second (and potentially slightly deeper) recursion in encode_interpolative and decode_interpolative into explicit iteration in case the compiler fails to optimise the tail recursion. Mon Dec 18 03:39:41 GMT 2006 Olly Betts <olly@survex.com> * matcher/: Add ExactPhrasePostList to handle the common special case where the window size of OP_PHRASE is equal to the number of terms in the phrase. Mon Dec 18 02:21:32 GMT 2006 Olly Betts <olly@survex.com> * tests/harness/testsuite.cc: Remove accidentally committed "NO PROBLEM" message when running under valgrind. Wed Dec 13 21:15:57 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * include/xapian/document.h: Another tweak to the documentation comment. Wed Dec 13 18:07:51 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * include/xapian/document.h: Improve documentation comment for Document::get_value(). Wed Dec 13 03:38:37 GMT 2006 Olly Betts <olly@survex.com> * include/xapian/enquire.h: In the description of the TradWeight class, replace reference to "Muscat 3.6" (meaningless to most people) with a generally useful explanation. Wed Dec 13 03:27:14 GMT 2006 Olly Betts <olly@survex.com> * docs/: svn:ignore XapianAdminNotes.html. Wed Dec 13 03:16:44 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: Yet more valgrind test improvements! Give up testing valgrind features if VALGRIND_COUNT_LEAKS isn't supported. Fix the common shortcut case to not try logging to /dev/null.<pid> so it can actually get used. Also don't use /dev/null in the --logfile case either as it will probably cause problems there too (though I don't have an old enough valgrind handy to verify this). Also report results of valgrind tests in a couple more places. Wed Dec 13 00:31:04 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * docs/Makefile.am: New rules for generating XapianAdminNotes.html from XapianAdminNotes.txt * configure.ac: Check for rst2html, and require it in maintainer mode. * HACKING: document rst2html as a requirement. Tue Dec 12 23:58:13 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * docs/XapianAdminNotes.txt: Add new document giving an introduction to Xapian concepts for system administrators. Tue Dec 12 21:33:38 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * docs/quartzdesign.html: Fix an unmatched bracket. Tue Dec 12 21:18:54 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * backends/quartz/quartz_alldocspostlist.cc: Fix memory leak discovered by valgrind in QuartzAllDocsPostList constructor: was failing to delete the Bcursor. Tue Dec 12 21:05:47 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * queryparser/queryparser.lemony: Fix parsing of queries of the form "+foo* bar", where no terms in the database match the wildcard "foo*", but bar does exist in the database. Previously, such queries would be equivalent to "bar". Now, they will match no documents. This required using a new "QpQuery" object internally so we can distinguish between deliberately "match nothing" queries, and empty lists of queries. * tests/api_anydb.cc,tests/queryparsertest.cc: Test above fix, and general parsing of wildcard queries with +terms, and test performing a match with a query resulting from a wildcard query which matches nothing. Tue Dec 12 21:05:15 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * tests/api_nodb.cc,api/omqueryinternal.cc,api/omquery.cc: Fix some copyright assertions I missed. Tue Dec 12 19:38:39 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * include/xapian/query.h,api/omqueryinternal.cc,api/omquery.cc: Add two static Query objects: Xapian::Query::MatchAll and Xapian::Query::MatchNothing, which match all documents and no documents in the database, respectively. (Equivalent to Query("") and Query()). Allow empty queries such as MatchNothing to be combined with other queries. Change methods of Xapian::Query::Internal to use pointers to query internals instead of references, so they can handle the NULL pointer internals of an empty query. * tests/api_nodb.cc: Test combining of MatchNothing queries with other queries with OP_AND and OP_OR. Tue Dec 12 20:01:58 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Fix valgrind better: --log-file-exactly isn't what is wanted, because it messes up if multiple processes are traced. Instead, use a temporary file instead of /dev/null, so we don't get the permission denied error. Tue Dec 12 19:15:20 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * configure.ac: Fix detection of valgrind for newer valgrind. My version of valgrind (valgrind-3.2.1-Debian, from ubuntu feisty) treats the parameter supplied to --log-file as a base path, and appends a process ID. This caused the test for the log-file option to fail with permission denied errors due to trying to write to files of the form '/dev/null.16098'. Solution - test for the --log-file-exactly option, and use that. Also, change configure.ac to display the result of the tests for valgrind; was displaying that it had found it, but not mentioning that it couldn't make it work. Tue Dec 05 21:12:12 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * include/xapian/query.h,api/omqueryinternal.cc: Fix query serialisation bug. Was failing to propagate "curpos" parameter across subqueries, resulting in incorrect serialisation of termpositions. * tests/internaltest.cc: Regression test for this bug. Tue Dec 05 01:34:07 GMT 2006 Olly Betts <olly@survex.com> * HACKING: Clarify how XAPIAN_DEBUG_FLAGS works. Tue Dec 05 01:23:12 GMT 2006 Olly Betts <olly@survex.com> * examples/quest.cc: Add "--stemmer" option to allow stemming language to be set, or stemming to be disabled. Sun Dec 03 00:34:27 GMT 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Add entries to the "unstem" map for prefixed boolean filters (e.g. type:html). Also don't corrupt non-ASCII characters in a prefixed boolean filter. Sat Nov 25 04:17:23 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_version.cc: The "my_fls()" change actually results in a different interpolative encoding in a few cases - the old encoding could vary between architectures so we have to change it. So bump the FLINT_VERSION and change the "flicklock" file to be "flintlock" as it should have been all along! Wed Nov 22 18:55:42 GMT 2006 Olly Betts <olly@survex.com> * xapian-config.in: Improve --version output so that help2man produces a better man page. Mon Nov 20 07:21:07 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Fix another off-by-one error (> should be >=). Mon Nov 20 07:12:11 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Fix off-by-one error in code which calls my_fls(). Sat Nov 18 08:32:08 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: "const static" -> "static const". Sat Nov 18 08:05:17 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Make decode_interpolative a member function of BitReader. Sat Nov 18 07:45:19 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Add highly optimised fls() implementation and make use of it. Thu Nov 16 04:22:36 GMT 2006 Olly Betts <olly@survex.com> * m4/xapian.m4: If XAPIAN_CONFIG wasn't specified and xapian-config wasn't found, see if the library seems to be present - if so give a different error message which suggests the user needs to install a -dev or -devel package. Thu Nov 16 02:11:34 GMT 2006 Olly Betts <olly@survex.com> * docs/install.html: omega tarball is now xapian-omega. Tue Nov 14 22:42:05 GMT 2006 Olly Betts <olly@survex.com> * tests/harness/Makefile: Add static Makefile so you can make all, check, and clean from the subdirectory. * tests/harness/Makefile.mk: Ship Makefile.mk and Makefile. Tue Nov 14 21:53:45 GMT 2006 Olly Betts <olly@survex.com> * tests/: Remove muscat36 temporary directories from svn:ignore property. Tue Nov 14 19:47:30 GMT 2006 Olly Betts <olly@survex.com> * configure.ac,tests/Makefile.am,tests/harness/Makefile.mk: Use non-recursive make to build the test harness, so it only gets built if "make check" is run. Tue Nov 14 19:43:48 GMT 2006 Olly Betts <olly@survex.com> * tests/quartztest.cc: Removed unused static function unlink_table. Tue Nov 14 17:41:26 GMT 2006 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.lemony, tests/queryparsertest.cc: Add QueryParser::FLAG_PURE_NOT to allow pure NOT queries to be enabled (they are now disabled by default). Tue Nov 14 04:19:40 GMT 2006 Olly Betts <olly@survex.com> * backends/database.cc,backends/flint/dir_contents, backends/quartz/dir_contents,docs/overview.html, docs/quartzdesign.html,docs/scalability.html: Make flint the default backend. Tue Nov 14 03:25:19 GMT 2006 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Xapian::Query("") now builds a query which matches all documents rather than throwing InvalidArgumentError so fix emptyquery1 test. Tue Nov 14 02:03:29 GMT 2006 Olly Betts <olly@survex.com> * api/omdocument.cc: Tweak OmDocumentTerm::add_position() so that adding position 0 to an empty termlist takes the shortcut. Mon Nov 13 05:46:15 GMT 2006 Olly Betts <olly@survex.com> * docs/queryparser.html: Document the new pure NOT feature. Mon Nov 13 05:24:30 GMT 2006 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,queryparser/queryparser.lemony, tests/queryparsertest.cc: Allow "pure NOT" queries - e.g. "NOT apples". Fixes bug #99. Mon Nov 13 04:54:04 GMT 2006 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Undo changes accidentally committed in last check-in. Mon Nov 13 04:47:57 GMT 2006 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,backends/Makefile.am,backends/database.cc, backends/muscat36/,configure.ac,docs/overview.html, docs/quartzdesign.html,docs/tests.html,include/xapian/dbfactory.h, include/xapian/version_h.cc,tests/api_anydb.cc,tests/apitest.cc, tests/harness/backendmanager.cc,tests/harness/backendmanager.h, tests/harness/index_utils.cc: Remove support for the old Muscat 3.6 backends. Mon Nov 13 04:02:27 GMT 2006 Richard Boulton <richard@lemurconsulting.com> * common/database.h,api/omdatabase.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/quartz/Makefile.am,backends/quartz/quartz_database.cc, backends/quartz/quartz_alldocspostlist.h, backends/quartz/quartz_alldocspostlist.cc, backends/flint/Makefile.am,backends/flint/flint_database.cc, backends/flint/flint_alldocspostlist.cc, backends/flint/flint_alldocspostlist.h: Implement posting lists which return a list of all documents in the database. Such a posting list is obtained by calling Xapian::Database::postlist_begin() with an empty term (ie, ""). Also, all Xapian::Database methods which take a termname now accept an empty term, and return appropriate values (ie, get_termfreq("") and get_collection_freq("") return the number of documents in the database, and term_exists("") returns true unless the database is empty). Fixes Bug #47. * docs/quartzdesign.html: Document the inefficiency of all-document postlists for Quartz. * tests/api_anydb.cc,tests/api_db.cc,tests/api_wrdb.cc: Add tests for all-document postlists, and for passing an empty term to all the applicable database methods. This defines the new tests allpostlist1, allpostlist2, emptyterm1, and emptyterm2. These tests currently skip for the remote backend where postlist_begin() isn't yet implemented. Mon Nov 13 02:06:03 GMT 2006 Olly Betts <olly@survex.com> * Merge in utf8 branch: Thu Sep 14 23:49:48 BST 2006 Olly Betts <olly@survex.com> * queryparser/,tests/queryparsertest.cc: Update to work with UTF-8. Stop normalising accents - the general sentiment seems to be firmly against it, and where it is still appropriate we should get the stemmers to do it. Sun Nov 12 22:38:56 GMT 2006 Olly Betts <olly@survex.com> * NEWS: Update from ChangeLog file in preparation for branching and merging. Sun Nov 12 19:32:01 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/quartz/quartz_database.cc: Fix replace_document() not to lose positional information for a document if it is replaced with itself with unmodified postings. * tests/api_wrdb.cc: Add testcase replace_document5 as regression test for the replace_document bug. * backends/remote/remote-database.cc: Fix RemoteDatabase::has_positions() to refetch the cached value if it might be out of date. Sun Nov 12 17:07:00 GMT 2006 Olly Betts <olly@survex.com> * docs/Makefile.am: Need to increase pool_size further still to build sourcedoc.pdf (2000000 now). Sun Nov 12 16:24:15 GMT 2006 Olly Betts <olly@survex.com> * HACKING: Add "update ReleaseNotes on wiki" to release checklist. Sun Nov 12 16:23:19 GMT 2006 Olly Betts <olly@survex.com> * xapian.spec.in: Remove "." from end of "Summary:". Package new man page for xapian-progsrv. Sun Nov 12 00:43:36 GMT 2006 Olly Betts <olly@survex.com> * docs/stemming.html: Update another "CVS" reference to say "SVN". Sun Nov 12 00:42:22 GMT 2006 Olly Betts <olly@survex.com> * docs/install.html: Update reference to "CVS" to say "SVN". Thu Nov 09 01:11:52 GMT 2006 Olly Betts <olly@survex.com> * HACKING: Reorder the release checklist a little. I've fixed update_website.sh to get the latest version from version.php, so now only version.php needs updating. Thu Nov 09 00:17:35 GMT 2006 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.ac: Update for 0.9.9. Wed Nov 08 20:54:31 GMT 2006 Olly Betts <olly@survex.com> * xapian.spec.in: Apply changes from Neal Becker to run "autoreconf --force" so that we don't set rpath for /usr/lib64, and add "libs" to %post and %postun. * AUTHORS: Thank Neal Becker. Wed Nov 08 04:27:17 GMT 2006 Olly Betts <olly@survex.com> * docs/tests.html: Update for "testsuite" -> "tests/harness". Mark paths and programs with <code>...</code>. Improve wording in a few places. Wed Nov 08 03:10:37 GMT 2006 Olly Betts <olly@survex.com> * bin/xapian-tcpsrv.cc: Need '#include "safeerrno.h"' for EADDRINUSE. Wed Nov 08 02:36:59 GMT 2006 Olly Betts <olly@survex.com> * bin/Makefile.am,bin/xapian-progsrv.cc,bin/xapian-tcpsrv.cc: xapian-progsrv now uses getopt for option processing, and we can now generate a man page using help2man. Fixes Bug #98. Wed Nov 08 01:47:19 GMT 2006 Olly Betts <olly@survex.com> * bin/xapian-tcpsrv.cc: If the port requested is in use, exit with code 69 (EX_UNAVAILABLE) which is useful if you're trying to automate launching of xapian-tcpsrv instances. * tests/harness/backendmanager.cc: If we can't start xapian-tcpsrv because the port is in use, try higher numbered ports. * tests/harness/testsuite.cc: Catch and report std::string exceptions. Tue Nov 07 22:40:58 GMT 2006 Olly Betts <olly@survex.com> * net/tcpclient.cc: Turn on TCP_NODELAY for the client too. Tue Nov 07 21:46:39 GMT 2006 Olly Betts <olly@survex.com> * net/tcpserver.cc: Turn on TCP_NODELAY for xapian-tcpsrv which increases throughput for the remote backend over tcp. Tue Nov 07 21:08:57 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_btreebase.cc: Add missing '#include'-s. Tue Nov 07 17:57:22 GMT 2006 Olly Betts <olly@survex.com> * backends/quartz/btree_base.cc: Add missing '#include'-s. Tue Nov 07 05:12:40 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.h,backends/flint/flint_table.cc, backends/quartz/btree.cc: Add missing '#include <vector>' which "utils.h" was implicitly pulling in. * common/utils.h: Move '#include "safefcntl.h"' to more logical place. Tue Nov 07 04:18:27 GMT 2006 Olly Betts <olly@survex.com> * common/safefcntl.h: Actually add the new file to SVN. Tue Nov 07 03:36:59 GMT 2006 Olly Betts <olly@survex.com> * HACKING,backends/flint/flint_io.h,backends/flint/flint_lock.cc, backends/flint/flint_lock.h,backends/muscat36/io_system.cc,common/, net/tcpserver.cc,tests/harness/backendmanager.cc, tests/harness/testsuite.cc: Create "safefcntl.h" as a replacement for <fcntl.h> instead of using "utils.h" for this purpose, since "utils.h" pulls in many other things we often don't want. * common/utils.cc,common/utils.h,net/progclient.cc: Move split_words to progclient.cc which is the only user of it. Rewrite it to not modified the string being split which risks being O(n^2). * net/progclient.cc,common/progclient.h: Pass std::string by const reference. * configure.ac: Fix DJGPP build (fork is present but always fails). Tue Nov 07 03:21:24 GMT 2006 Olly Betts <olly@survex.com> * matcher/biaspostlist.h: Don't need '#include "utils.h"'. Tue Nov 07 01:15:14 GMT 2006 Olly Betts <olly@survex.com> * common/c_strtod.cc: Remove file which was checked in but is unused! Mon Nov 06 15:24:02 GMT 2006 Olly Betts <olly@survex.com> * tests/internaltest.cc: Disable serialiselength1 and serialisedoc1 when the remote backend is disabled. Mon Nov 06 15:23:40 GMT 2006 Olly Betts <olly@survex.com> * tests/Makefile.am: Fix typo in recent check-in. Mon Nov 06 01:43:37 GMT 2006 Olly Betts <olly@survex.com> * tests/Makefile.am: Need "harness" in "$(srcdir)" for VPATH builds to work. Sun Nov 05 19:55:26 GMT 2006 Olly Betts <olly@survex.com> * backends/quartz/Makefile.am: Remove explicit dependency of libbtreecheck.la on libxapian.la. We always link in libxapian.la and the explicit dependency makes it hard to build things in a sane order. Sun Nov 05 19:29:17 GMT 2006 Olly Betts <olly@survex.com> * net/tcpserver.cc: Don't define on_SIGCHLD() unless we'll use it. * tests/harness/backendmanager.cc: Use a proper signal handler for SIGCHLD if we have waitpid() - POSIX leaves the semantics of SIG_IGN on SIGCHLD unspecified. Sun Nov 05 19:11:36 GMT 2006 Olly Betts <olly@survex.com> * backends/quartz/Makefile.am,backends/quartz/btreecheck.cc, backends/quartz/btreecheck.h,bin/Makefile.am,tests/Makefile.am, tests/harness/: Use _exit(0) instead of exit(0) in the child process which closes the pipe to xapian-tcpsrv in BackendManager since we don't want to call atexit functions from the child. Move btreecheck stuff into backends/quartz. Sun Nov 05 17:12:05 GMT 2006 Olly Betts <olly@survex.com> * Makefile.am,bin/Makefile.am,configure.ac,tests/Makefile.am, tests/harness/Makefile.am,tests/harness/dir_contents,testsuite/: Move the testsuite harness from "testsuite/" to "tests/harness/". Sun Nov 05 16:36:56 GMT 2006 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc,testsuite/backendmanager.h: Now needs <stdio.h>. Set SIG_IGN for SIG_CHLD so we don't get zombie child processes. Sun Nov 05 03:53:24 GMT 2006 Olly Betts <olly@survex.com> * tests/apitest.cc: Make backendmanager static. Sun Nov 05 03:51:36 GMT 2006 Olly Betts <olly@survex.com> * bin/xapian-tcpsrv.cc: Output "Listening..." once the socket is open and read for connections. * testsuite/backendmanager.cc: Use popen() to run xapian-tcpsrv and wait for "Listening..." before returning rather than just sleeping for 1 second and hoping that's enough. Sun Nov 05 02:54:23 GMT 2006 Olly Betts <olly@survex.com> * tests/api_db.cc: Remove unnecessary inclusion of backendmanager.h. Fri Nov 03 02:09:26 GMT 2006 Olly Betts <olly@survex.com> * HACKING: Update debian packaging checklist. Fri Nov 03 00:57:35 GMT 2006 Olly Betts <olly@survex.com> * docs/Makefile.am: Building sourcedoc.pdf needs a larger pool_size now. Thu Nov 02 19:12:34 GMT 2006 Olly Betts <olly@survex.com> * AUTHORS: Updated. Thu Nov 02 15:41:46 GMT 2006 Olly Betts <olly@survex.com> * HACKING,NEWS,configure.ac: Update for 0.9.8. Thu Nov 02 15:20:05 GMT 2006 Olly Betts <olly@survex.com> * PLATFORMS: Update from tinderbox. Thu Nov 02 11:53:53 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: GCC's -Wendif-labels is enabled by default on versions which support it, which simplifies our tests. Thu Nov 02 00:22:44 GMT 2006 Olly Betts <olly@survex.com> * backends/remote/dir_contents: Update. Wed Nov 01 15:33:12 GMT 2006 Olly Betts <olly@survex.com> * common/omstringstream.h: Fix our implementation of om_ostringstream to work with OmTime. Wed Nov 01 03:27:36 GMT 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Don't require a prefixed boolean term to start with an alphanumeric - allow the same set of characters as we do for the second and subsequent characters. Sat Oct 28 04:12:13 BST 2006 Olly Betts <olly@survex.com> * bin/quartzcheck.cc: Add catch for unknown exceptions. Sat Oct 28 03:49:19 BST 2006 Olly Betts <olly@survex.com> * HACKING,configure.ac,docs/Makefile.am,docs/doxygen_api_conf.in, docs/doxygen_full_conf.in: Produce a PDF for apidoc rather than PostScript, since the PDF is smaller, and easier to view for most users. Use pdflatex to generate the PDF directly rather than going via a DVI file. This also avoids problems on some Linux distros where latex is a symlink to pdfelatex (bug#81, bug#95). Thu Oct 26 22:12:20 BST 2006 Richard Boulton <richard@lemurconsulting.com> * bin/quartzcheck.cc: Catch (and display) any exceptions which are of type "const char *error". btreecheck.cc raises an exception of this type if a btree error is found, and this avoids quartzcheck dying quite so horribly in this case. Wed Oct 25 23:30:26 BST 2006 Olly Betts <olly@survex.com> * configure.ac: -Wendif-labels is new in GCC 3.3. * configure.ac: Revert accidental change which turned on -Werror in non-maintainer builds. * configure.ac: Avoid non-portable use of double quotes in double-quoted backticks. Wed Oct 25 01:19:24 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Redhat's GCC 2.96 doesn't support -Wundef even though real GCC version before and after it do! Also, use -Wshadow and -Wendif-labels even when not in maintainer mode. Tue Oct 24 04:17:58 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc,bin/quartzcheck.cc: Eliminate a couple of variables whose value is never used. Tue Oct 24 00:31:25 BST 2006 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Only force a flush on WritableDatabase::allterms_begin() if there are actually pending changes. Mon Oct 23 23:49:52 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Only force a flush on WritableDatabase::allterms_begin() if there are actually pending changes. Mon Oct 23 02:24:12 BST 2006 Olly Betts <olly@survex.com> * configure.ac: When checking if we need -lm, don't use a constant argument to log as the compiler might simply evaluate the whole expression at compile time. Sat Oct 21 20:42:52 BST 2006 Olly Betts <olly@survex.com> * HACKING: Mention automake 1.10 is out but we've not tested it yet. Sat Oct 21 20:39:57 BST 2006 Olly Betts <olly@survex.com> * HACKING: Add entries to release checklist: make sure new API methods are wrapped by the bindings, and that bug submitters are thanked. Fri Oct 20 13:56:50 BST 2006 Richard Boulton <richard@lemurconsulting.com> * backends/flint/flint_io.cc: Fix compilation on windows (needs to #include "safewindows.h" to get definition of SSIZE_T). Tue Oct 17 02:16:37 BST 2006 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc: Fix compilation when valgrind is detected by configure. Thu Oct 12 13:30:05 BST 2006 Olly Betts <olly@survex.com> * xapian.spec.in: Package xapian-progsrv. Thu Oct 12 00:49:47 BST 2006 Olly Betts <olly@survex.com> * HACKING: Note that on Debian, tetex-extra is needed for fancyhdr.sty. * HACKING: Note that dch can be used to update debian/changelog. Wed Oct 11 23:35:08 BST 2006 Olly Betts <olly@survex.com> * docs/Makefile.am: If running latex on refman.ps fails, cat refman.log since that is likely to show what failed. Tue Oct 10 17:24:00 BST 2006 Olly Betts <olly@survex.com> * NEWS: Bump release date. Sun Oct 08 21:41:04 BST 2006 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.ac: Update for 0.9.7. Sun Oct 08 10:06:51 BST 2006 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Use lseek() to skip existing valgrind output instead of repeated calls to read. Handle the old valgrind naming convention for log files. Sun Oct 08 09:35:59 BST 2006 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix a couple of typos in comments. Sun Oct 08 05:36:36 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Disable probing and short-cut tests for a FORTRAN compiler. We don't use one, but current libtool versions always check for it regardless. Sat Oct 07 21:19:35 BST 2006 Olly Betts <olly@survex.com> * configure.ac,tests/runtest.in,testsuite/backendmanager.cc, testsuite/testsuite.cc: Fix testsuite harness to show valgrind output when a test fails (when running under valgrind in verbose mode). This probably stopped working due to changes in valgrind 3. * testsuite/backendmanager.cc: Run xapian-tcpsrv under valgrind if apitest is. Fri Oct 6 18:27:13 BST 2006 Richard Boulton <richard@lemurconsulting.com> * tests/runtest.in: export $LIBTOOL, $VALGRIND and $VG_LOG_FD. * testsuite/backendmanager.cc: If $LIBTOOL, $VALGRIND and $VG_LOG_FD are set, run progsrv under valgrind. Fixes Bug #94. * net/remoteserver.cc: Use an AutoPtr to hold the unserialised query, so it gets deleted if an exception is thrown. * api/omqueryinternal.cc: Use AutoPtr in one place, and a try-catch in another, to ensure that partially unserialised queries get deleted if exceptions are thrown. * tests/internaltest.cc: Add copyright notices to file for recent change. * AUTHORS: Add myself as a current developer, now that I've started committing to the core again. Fri Oct 06 17:44:21 BST 2006 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Fix memory leak in query unserialisation. Fri Oct 6 17:34:59 BST 2006 Richard Boulton <richard@lemurconsulting.com> * tests/runtest.in: Cache result of test for $VG_LOG_FD if we find the new option. Saves 2 seconds for each invocation on my machine, which is half the time when running just a single simple test. Fri Oct 6 16:35:46 BST 2006 Richard Boulton <richard@lemurconsulting.com> * tests/internaltest.cc: Check serialisation and unserialisation of Query objects. Currently fails under valgrind due to a memory leak somewhere in the unserialisation code. Thu Oct 05 14:13:55 BST 2006 Olly Betts <olly@survex.com> * include/xapian/error.h: Only enable the SWIG visibility hook when using GCC 4 or later. Wed Oct 04 20:32:43 BST 2006 Olly Betts <olly@survex.com> * m4/xapian.m4: Remove overquoting. Wed Oct 4 13:28:09 BST 2006 Richard Boulton <richard@lemurconsulting.com> * common/utils.h: MSVC seems to #define open. However, the workaround for this problem implemented for old versions of solaris doesn't work for windows, so avoid applying the fix for windows, and just #undef open. (Windows seems to define some open() functions, as well as #defining open!) Wed Oct 4 13:26:18 BST 2006 Richard Boulton <richard@lemurconsulting.com> * net/serialise.cc: Use "unsigned char" instead of "char" when serialising lengths, to avoid problems on platforms where char is signed (eg, windows). Wed Oct 04 12:25:51 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_version.cc: Remove "100" from start of file so it actually compiles. Mon Oct 02 13:57:56 BST 2006 Olly Betts <olly@survex.com> * xapian-config.in: Fix typo - "@libdir" should be "@libdir@". This would lead to -L/usr/lib not being pruned, which is really just a cosmetic problem (the typo was introduced in 0.9.3). Thu Sep 28 02:00:05 BST 2006 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Reserve the right number of entries in the subrsets vector. Fri Sep 22 07:48:32 BST 2006 Olly Betts <olly@survex.com> * common/serialise-double.cc: Fix warning with aCC. Fri Sep 22 04:43:06 BST 2006 Olly Betts <olly@survex.com> * HACKING: Expand note on _GLIBCXX_DEBUG; Now using autoconf 2.60 for snapshots and releases; Now using a libtool patch which improves support for -library=stlport4 with Sun's C++; Give URL to Alexandre Duret-Lutz's autotools tutorial, which is much more up-to-date than the "goat book". Fri Sep 22 04:42:08 BST 2006 Olly Betts <olly@survex.com> * common/serialise-double.cc: Fix a few compiler warnings. Fri Sep 22 04:29:18 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc: Retry on EINTR from fcntl or waitpid. Fri Sep 22 03:39:12 BST 2006 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: Only check _GLIBCXX_DEBUG for GCC 3.4 and later (which are the versions which support it). Fri Sep 22 03:37:02 BST 2006 Olly Betts <olly@survex.com> * bin/xapian-tcpsrv.cc: Report errno if we catch a Xapian::Error which has it set. Fri Sep 22 03:30:25 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Turn on -Wportability to help ensure our Makefile.am's are written in a portable way. Fri Sep 22 03:29:25 BST 2006 Olly Betts <olly@survex.com> * tests/runtest.in: Turn on GLIBCXX_FORCE_NEW when running tests under valgrind. Tue Sep 19 06:28:47 BST 2006 Olly Betts <olly@survex.com> * tests/internaltest.cc: Check that the the destructor on a temporary object gets called at the correct time (Sun C++ deliberately gets this wrong by default). Tue Sep 19 04:32:04 BST 2006 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Revert change to Xapian::Weight's copy constructor because it prevents Omega from compiling. Tue Sep 19 04:01:14 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Make the argument of log() a double to avoid potential compiler warnings. Mon Sep 18 22:58:19 BST 2006 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Xapian::Weight's copy constructor should be private not protected (direct copying isn't allowed). Mon Sep 18 07:03:51 BST 2006 Olly Betts <olly@survex.com> * configure.ac: We reportedly need "-lm" to get maths functions on some versions of Sun's C++ compiler. Sat Sep 16 12:08:25 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Sun's C++ compiler implements non-standards-conforming lifetimes for temporary objects (for "backwards compatibility" with old Sun C++ specific code). We don't care about such code, so always pass "-features=tmplife" for Sun C++. Sat Sep 16 03:04:57 BST 2006 Olly Betts <olly@survex.com> * bin/xapian-progsrv.cc: Oops, fix compilation error. Sat Sep 16 02:13:22 BST 2006 Olly Betts <olly@survex.com> * bin/xapian-progsrv.cc: Fix messages send by xapian-progsrv if an exception is thrown while opening the database. Fri Sep 15 06:29:55 BST 2006 Olly Betts <olly@survex.com> * tests/internaltest.cc: 1/DBL_MAX may be less than DBL_MIN but on platforms like x86 which hold results in registers with extra precision, this is still representable, but not reliably so the test sometimes fails. Fri Sep 15 02:40:52 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_btreebase.cc: Avoid copying beyond the end of the bitmap block. Thu Sep 14 02:01:58 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_version.cc: Fix warning from GCC 4.1.0. Wed Sep 13 18:12:43 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_io.cc,backends/flint/flint_io.h, backends/flint/flint_version.cc: Fix compiler warnings. Wed Sep 13 06:09:40 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_version.cc,backends/flint/flint_version.h: Actually commit the new files! Wed Sep 13 05:21:04 BST 2006 Olly Betts <olly@survex.com> * backends/flint/,common/utils.h: Rewrite some of flint's low level IO functions, and the "iamflint" handling class. Tue Sep 12 20:22:57 BST 2006 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc: Fix to compile. Tue Sep 12 18:56:16 BST 2006 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remote-database.h: Fix bug in remote backend which incorrectly returned an empty MSet under certain circumstances! Tue Sep 12 11:51:31 BST 2006 Olly Betts <olly@survex.com> * matcher/msetcmp.cc: "static inline" -> "inline" since the static is superfluous and Sun's C++ warns. Mon Sep 11 23:42:28 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Check $CXX not $CC to identify which C++ compiler we have. Mon Sep 11 16:32:37 BST 2006 Olly Betts <olly@survex.com> * api/omenquire.cc,api/vectortermlist.h, backends/flint/flint_positionlist.cc,configure.ac, net/remoteconnection.cc,net/tcpclient.cc, queryparser/queryparser.lemony,tests/api_anydb.cc,tests/api_db.cc, tests/api_nodb.cc: I've discovered that -library=stlport4 puts Sun's compiler into an "ANSI C++ compliant" mode, so do that automatically in configure and throw away all the annoying special bits of alternative code we'd accumulated just for this one compiler. Mon Sep 11 16:01:20 BST 2006 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Speed up deldoc4 when run in verbose mode - some stringstream implementations are very inefficient when the string grows long. Mon Sep 11 05:30:29 BST 2006 Olly Betts <olly@survex.com> * HACKING: Add some advice regarding debugging using -D_GLIBCXX_DEBUG, valgrind, and gdb. Sun Sep 10 02:24:47 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Fix last check-in to actually work. Sat Sep 09 04:19:44 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Ensure that if _GLIBCXX_DEBUG has been specified that it also passed when generating version.h. Sat Sep 09 04:01:40 BST 2006 Olly Betts <olly@survex.com> * tests/internaltest.cc: Give more useful output should the double serialisation test fail. Sat Sep 09 03:19:20 BST 2006 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: Add a check that _GLIBCXX_DEBUG is set compatibly if we're compiling with GNU C++. Sat Sep 09 02:55:38 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_modifiedpostlist.cc: Fix potential access to iterator which has already reached its end. Fri Sep 08 04:05:28 BST 2006 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/,matcher/bm25weight.cc, matcher/tradweight.cc,net/,tests/internaltest.cc: Split the double serialisation code off into its own file - it is used by BM25Weight and TradWeight, so it needs to be compiled in even when the remote backend is disabled. Thu Sep 07 00:09:41 BST 2006 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc: Discard stderr from xapian-tcpsrv so we don't get "write error" messages appearing in the testsuite output when we've just closed the connection at the client side. Tue Sep 05 21:07:40 BST 2006 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Fix warning when remote backend is disabled. Tue Sep 05 20:58:52 BST 2006 Olly Betts <olly@survex.com> * docs/Makefile.am: Add extra dependencies so that parallel make doesn't try to run latex twice simultaneously. Tue Sep 05 20:58:13 BST 2006 Olly Betts <olly@survex.com> * AUTHORS: Updated. Tue Sep 05 20:56:50 BST 2006 Olly Betts <olly@survex.com> * Makefile.am: Fix typo. Tue Sep 05 03:23:14 BST 2006 Olly Betts <olly@survex.com> * docs/queryparser.html,queryparser/queryparser.lemony, tests/queryparsertest.cc: Implement "ADJ" operator - like "NEAR" except the terms must appear in matching documents in the same order as in the query. Tue Sep 05 03:19:12 BST 2006 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/,docs/remote_protocol.html, matcher/bm25weight.cc,matcher/tradweight.cc,net/remoteserver.cc, net/serialise.cc,tests/internaltest.cc: Change how doubles are serialised by TradWeight, BM25Weight, and in the remote backend protocol. The new encoding allows us to transfer any double value which can be represented by both machines precisely and compactly. Tue Sep 05 02:27:36 BST 2006 Olly Betts <olly@survex.com> * docs/queryparser.html,queryparser/queryparser.lemony, tests/queryparsertest.cc: Allow a distance to be specified for NEAR - e.g. "cats NEAR/3 dogs" (bug#92). Fri Sep 01 00:29:10 BST 2006 Olly Betts <olly@survex.com> * backends/remote/remote-database.cc,common/remote-database.h: Fix RemoteDatabase::reopen() to not be const so it actually overrides the virtual method it is supposed to. Thu Aug 31 21:40:53 BST 2006 Olly Betts <olly@survex.com> * testsuite/backendmanager.h: Remove unneeded BackendManager:: qualifications. Thu Aug 31 17:08:16 BST 2006 Olly Betts <olly@survex.com> * matcher/msetpostlist.cc,matcher/msetpostlist.h: Move #include "omenquireinternal.h" into the header to fix compilation error with older versions of GCC. Thu Aug 31 16:38:33 BST 2006 Olly Betts <olly@survex.com> * PLATFORMS: Added success report for Nexenta (alpha 5). Wed Aug 30 23:41:08 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Fix generation of version.h to work with Solaris sed. Sat Aug 26 15:28:22 BST 2006 Olly Betts <olly@survex.com> * docs/index.html: Add links to the wiki. Sun Jul 16 03:48:26 BST 2006 Olly Betts <olly@survex.com> * common/Makefile.am: Ship remoteprotocol.h. Sun Jul 16 02:23:54 BST 2006 Olly Betts <olly@survex.com> * common/remote-database.h: Add new file I missed in the previous commit. Sun Jul 16 01:58:25 BST 2006 Olly Betts <olly@survex.com> * Makefile.am,api/,backends/Makefile.am,backends/database.cc, backends/dbfactory_remote.cc,backends/flint/flint_database.cc, backends/flint/flint_termlist.cc,backends/flint/flint_termlist.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/multi/multi_termlist.cc,backends/multi/multi_termlist.h, backends/net/,backends/quartz/quartz_termlist.cc, backends/quartz/quartz_termlist.h,backends/remote/,bin/Makefile.am, bin/xapian-progsrv.cc,bin/xapian-tcpsrv.cc,common/,configure.ac, docs/remote_protocol.html,include/xapian/,matcher/,net/,tests/, testsuite/backendmanager.cc,testsuite/backendmanager.h: Rewrite most of the remote backend. It now supports most operations which a local database does (including writing!), the protocol used is more compact, and a number of layers of classes have been eliminated and the sequences of method calls simplified, so the code should be easier to understand and maintain despite doing more. A number of bugs have been fixed in the process. Sat Jul 15 05:10:38 BST 2006 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Doesn't need <iostream>. Sat Jul 15 01:06:27 BST 2006 Olly Betts <olly@survex.com> * README: Add link to the wiki. Tweak wording. Fri Jul 14 15:21:39 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Note in error message that dot is in graphviz. Fri Jul 14 15:17:36 BST 2006 Olly Betts <olly@survex.com> * docs/overview.html: Add discussion of uses of terms vs values. Fri Jul 14 15:01:04 BST 2006 Olly Betts <olly@survex.com> * docs/overview.html: Rewrite the section on Xapian::Document to remove some very out-of-date information and make it clearer. Tue Jul 11 18:41:07 BST 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix problem I believe was introduced by previous fix. * tests/queryparsertest.cc: Add regression test and some additional related test cases. Tue Jul 11 03:32:48 BST 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix bug in how we handle prefixed quoted phrases and prefixed brackets. * tests/queryparsertest.cc: Add regression tests. Mon Jul 10 23:17:58 BST 2006 Olly Betts <olly@survex.com> * include/xapian/database.h: Note that automatically allocated document IDs don't reuse IDs from deleted document. Wed Jul 05 01:06:35 BST 2006 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Tweak whitespace. Mon Jun 26 23:56:02 BST 2006 Olly Betts <olly@survex.com> * PLATFORMS: Added success reports for MSVC and sparc linux. Sat Jun 17 02:01:35 BST 2006 Olly Betts <olly@survex.com> * Makefile.am,tests/Makefile.am: Tweak new check-* rules to be more portable and robust. Sun Jun 11 23:29:48 BST 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Fix parsing of loved and hated prefixed phrases and bracketted expressions. Fix handling of stopwords in boolean expressions. Don't ignore a stopword if it's the only query term. Add regression tests for all these cases. Fri Jun 09 15:21:07 BST 2006 Olly Betts <olly@survex.com> * docs/queryparser.html: Add pointer to set_database when describing FLAG_WILDCARD. Fri Jun 09 13:51:03 BST 2006 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Add note that FLAG_WILDCARD requires you to call set_database. Fri Jun 09 13:49:34 BST 2006 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Don't compile query serialisation if the remote backend is disabled. Fri Jun 09 01:48:25 BST 2006 Olly Betts <olly@survex.com> * api/omdocument.cc,tests/api_nodb.cc: add_value failed to replace an existing value with the same number, contrary to what the documentation says (bug #82). Thu Jun 08 21:36:54 BST 2006 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Don't fetch the document data when fetching the value to sort on. Simple benchmarking showed this to speed up sort by value by a factor of between 3 and 9! Sun Jun 04 17:36:01 BST 2006 Olly Betts <olly@survex.com> * backends/flint/: Remove forced flush when iterating the posting list of a term which has modified posting pending. Sat Jun 03 21:38:43 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_termlist.cc, backends/quartz/quartz_database.cc,tests/api_wrdb.cc: We can't flush during a transaction, which means that we can't use flush to avoid having to handle corner cases (like deleting a document right after adding it before it's been flushed) so handle corner cases properly (except for postlist_begin() and allterms_begin() which are harder - these now throw UnimplementedError at least...) * backends/flint/flint_database.cc,backends/quartz/quartz_database.cc: replace_document(did, doc) was double-incrementing the "changes" counter when document did didn't exist - fixed. Sat Jun 03 17:53:41 BST 2006 Olly Betts <olly@survex.com> * HACKING: Document "make check-flint" and "make check-quartz". Sat Jun 03 17:49:25 BST 2006 Olly Betts <olly@survex.com> * Makefile.am,tests/Makefile.am: Added make targets "check-flint" and "check-quartz" which run the subset of tests which test the flint and quartz backends respectively. Sat Jun 03 04:03:00 BST 2006 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/database.cc, backends/flint/flint_database.cc,backends/flint/flint_database.h, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h,backends/quartz/dir_contents, backends/quartz/quartz_database.cc,backends/quartz/quartz_database.h, common/database.h,include/xapian/database.h,tests/apitest.cc: Rework transactions to support "unflushed" transactions, and so they work with quartz as well as with flint. Sat Jun 03 03:18:22 BST 2006 Olly Betts <olly@survex.com> * HACKING: Document "make check-remote". Sat Jun 03 00:23:46 BST 2006 Olly Betts <olly@survex.com> * backends/database.cc,backends/flint/dir_contents, backends/flint/flint_database.cc,backends/flint/flint_database.h, backends/quartz/dir_contents,common/database.h, include/xapian/database.h,tests/: Implement transactions for flint. Sat Jun 03 00:14:37 BST 2006 Olly Betts <olly@survex.com> * tests/apitest.cc: Run tests on flint if flint is enabled, rather than if quartz is enabled. Sun May 28 23:01:45 BST 2006 Olly Betts <olly@survex.com> * common/omtime.h: Add operator+ and operator+= with argument of type Xapian::timeout. Sun May 28 22:02:30 BST 2006 Olly Betts <olly@survex.com> * include/xapian/errorhandler.h: Fix typos in private assignment operator and copy ctor (Error -> ErrorHandler!) Sun May 28 21:54:05 BST 2006 Olly Betts <olly@survex.com> * Makefile.am,tests/Makefile.am: Add "check-remote" target which runs the subset of tests which test the remote backend. Thu May 25 16:06:06 BST 2006 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/database.cc,common/database.h: Merge Xapian::Internal::open_database() into the Xapian::Database ctor which calls it; merge Xapian::Internal::open_writable_database() into the Xapian::WritableDatabase ctor which calls it. Wed May 24 08:28:40 BST 2006 Olly Betts <olly@survex.com> * net/socketserver.cc: OmLineBuf -> OmSocketLineBuf. Wed May 24 08:27:22 BST 2006 Olly Betts <olly@survex.com> * common/socketserver.h: OmLineBuf -> OmSocketLineBuf. Wed May 24 07:50:51 BST 2006 Olly Betts <olly@survex.com> * common/Makefile.am,common/omlinebuf.h,common/socketcommon.h, net/Makefile.am,net/omlinebuf.cc,net/socketcommon.cc: Merge OmLineBuf into OmSocketLineBuf. * common/Makefile.am: Fix netutils.cc to be conditionally included when the remote backend is enabled, not the quartz backend! Tue May 23 18:59:09 BST 2006 Olly Betts <olly@survex.com> * api/Makefile.am,api/errorhandler.cc,api/omerror.cc, include/xapian/error.h,include/xapian/errorhandler.h: Redo the Xapian::Error and Xapian::ErrorHandler classes. The new versions have better, clearer documentation comments and are cleaner internally. * include/xapian/error.h: Add hook to allow SWIG bindings to be built using GCC's visibility support. Tue May 23 10:08:02 BST 2006 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Remove superfluous '#include <autoptr.h>'. Tue May 23 09:42:52 BST 2006 Olly Betts <olly@survex.com> * backends/Makefile.am,backends/database.cc, backends/dbfactory_remote.cc,backends/net/net_database.cc, backends/net/net_termlist.cc,backends/net/net_termlist.h,common/, include/xapian/dbfactory.h,matcher/,net/socketclient.cc, net/socketcommon.cc: Eliminate the NetClient class by merging it into NetDatabase. Mon May 22 08:51:16 BST 2006 Olly Betts <olly@survex.com> * common/omtime.h: Fix OmTime::operator> which failed to return false if the seconds were strictly less but the microsecond fraction was more. Sun May 21 11:53:29 BST 2006 Olly Betts <olly@survex.com> * Makefile.am,bin/Makefile.am,docs/Makefile.am,examples/Makefile.am: Make use of the dist_ prefix to avoid having to list files in EXTRA_DIST as well as in *_DATA and man_MANS. Sun May 21 05:55:00 BST 2006 Olly Betts <olly@survex.com> * api/dir_contents,getopt/dir_contents,queryparser/dir_contents: Add missing dir_contents files. Sun May 21 05:45:44 BST 2006 Olly Betts <olly@survex.com> * tests/remotetest.cc: Remove unnecessary "#include <sys/wait.h>". Sat May 20 11:02:43 BST 2006 Olly Betts <olly@survex.com> * docs/Makefile.am: doxygen_api_conf and doxygen_full_conf are generated, so aren't in srcdir! Sat May 20 10:14:52 BST 2006 Olly Betts <olly@survex.com> * docs/Makefile.am: automake adds suitable rules for rebuilding doxygen_api_conf and doxygen_source_conf, so remove our less accurate versions. Fri May 19 14:43:47 BST 2006 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/doxygen_api_conf.in,docs/doxygen_full_conf.in: Remove "XAPIAN_DEPRECATED" from generated documentation. * docs/Makefile.am: Fix dependencies for regenerating the doxygen documentation. Fri May 19 08:12:51 BST 2006 Olly Betts <olly@survex.com> * docs/scalability.html: quartzcompact and xapian-compact now allow you to set the blocksize, so there's no need to use copydatabase if you want to migrate a database to a larger blocksize. Mention gmane. Other minor tweaks. Fri May 19 07:52:23 BST 2006 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Add --blocksize option to allow the blocksize to be set (default is 8K as before.) Fri May 19 07:47:13 BST 2006 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Add --blocksize option to allow the blocksize to be set (default is 8K as before.) Thu May 18 11:24:17 BST 2006 Olly Betts <olly@survex.com> * api/vectortermlist.h: Whitespace tweak. Tue May 16 10:09:53 BST 2006 Olly Betts <olly@survex.com> * HACKING: Update details of the debian stable backport version numbering scheme. Tue May 16 06:55:14 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Remove unused variable from snprintf testing code. Tue May 16 04:32:34 BST 2006 Olly Betts <olly@survex.com> * HACKING: Expand on the debian package building checklist. Tue May 16 04:32:07 BST 2006 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Note that "set_sort_by_relevance" is the default setting. Mon May 15 06:59:01 BST 2006 Olly Betts <olly@survex.com> * HACKING: Update debian package building checklist. Mon May 15 03:53:53 BST 2006 Olly Betts <olly@survex.com> * PLATFORMS: Update one more result before the actual release. Mon May 15 02:15:18 BST 2006 Olly Betts <olly@survex.com> * NEWS,configure.ac: Updated for 0.9.6. Mon May 15 01:35:33 BST 2006 Olly Betts <olly@survex.com> * PLATFORMS: Updated in preparation for the next release. Sun May 14 19:05:37 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_lock.h: Added workaround for newlib header bug. Sat May 13 07:04:14 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Fix snprintf tests. Sat May 13 04:52:53 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Tweak version.h generation to cope with CXXCPP putting carriage returns into its output as can happen on cygwin. Fri May 12 21:49:33 BST 2006 Olly Betts <olly@survex.com> * HACKING: Update with the libtool patches we're now using. Fri May 12 21:43:02 BST 2006 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: Trim trailing whitespace. Fri May 12 20:43:47 BST 2006 Olly Betts <olly@survex.com> * configure.ac,include/xapian/version_h.cc: Replace @@ with " instead of @, so we can write @deprecated. Fix more compilation problems. Fri May 12 19:00:08 BST 2006 Olly Betts <olly@survex.com> * include/xapian/version_h.cc: Fix compilation problem. Fri May 12 01:19:14 BST 2006 Olly Betts <olly@survex.com> * api/version.cc,include/xapian/version_h.cc: Rename Xapian::xapian_version_string() and companions to Xapian::version_string(). Keep the old functions as aliases which are marked as deprecated. Wed May 10 18:25:59 BST 2006 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Remove bogus documentation for a parameter which doesn't exist. Tue May 09 19:17:23 BST 2006 Olly Betts <olly@survex.com> * bin/Makefile.am: Remove trailing whitespace. * bin/xapian-compact.cc: Fix renaming of "iamflint.tmp" for MS Windows where you can't rename an open file. Tue May 09 15:57:26 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Fix reversed conditional in test for snprintf (which affects cygwin). Mon May 01 21:49:46 BST 2006 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add another prefix testcase to improve coverage. Sat Apr 29 20:16:46 BST 2006 Olly Betts <olly@survex.com> * docs/remote_protocol.html: Document keep-alive messages. Thu Apr 13 14:49:48 BST 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Add rules to handle a boolean filter with a "+" in front (such as +site:xapian.org). Wed Apr 12 18:42:31 BST 2006 Olly Betts <olly@survex.com> * PLATFORMS: More updates for 0.9.5. Tue Apr 11 19:56:16 BST 2006 Olly Betts <olly@survex.com> * matcher/Makefile.am: Need to add "-I${top_builddir}/include" to INCLUDES so that xapian/version.h is found. Tue Apr 11 19:24:59 BST 2006 Olly Betts <olly@survex.com> * backends/database.cc,configure.ac,include/xapian/version_h.cc, matcher/multimatch.cc,matcher/stats.cc,tests/apitest.cc, testsuite/backendmanager.cc,testsuite/backendmanager.h: Eliminate XAPIAN_BUILD_BACKEND_* from config.h and just use XAPIAN_HAS_*_BACKEND from xapian/version.h instead. Tue Apr 11 18:32:52 BST 2006 Olly Betts <olly@survex.com> * include/Makefile.am: Add xapian/version.h.timestamp as a dependency on all-local so that xapian/version.h actually gets regenerated when required. Tue Apr 11 17:52:24 BST 2006 Olly Betts <olly@survex.com> * api/omenquire.cc,backends/flint/flint_btreebase.cc, backends/quartz/btree_base.cc,common/utils.h,configure.ac: Disable MSVC warning 4800 (on int to bool conversions) in config.h and then we can remove the "fixes" elsewhere. Tue Apr 11 16:28:01 BST 2006 Olly Betts <olly@survex.com> * configure.ac: Simpler check for VALGRIND being set to empty value. Tue Apr 11 01:04:32 BST 2006 Olly Betts <olly@survex.com> * PLATFORMS: Add a summary. Tue Apr 11 00:45:55 BST 2006 Olly Betts <olly@survex.com> * PLATFORMS: Updates from boxes which were down when I did the release. Mon Apr 10 17:06:46 BST 2006 Olly Betts <olly@survex.com> * api/omenquire.cc,backends/flint/flint_btreebase.cc, backends/flint/flint_utils.h,backends/quartz/btree_base.cc, backends/quartz/quartz_utils.h,common/omassert.h: Fix more MSVC7 warnings (I spoke too soon). Mon Apr 10 15:56:52 BST 2006 Olly Betts <olly@survex.com> * include/xapian/query.h: Another MSVC7 warning fix. Should be free of warnings now. Mon Apr 10 14:46:34 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_utils.h, backends/quartz/quartz_utils.h,bin/quartzcompact.cc, bin/xapian-compact.cc,common/omdebug.h,common/utils.h, include/xapian/query.h,languages/header.h,matcher/multimatch.cc: Fix assorted MSVC7 warnings. Sun Apr 09 04:56:09 BST 2006 Olly Betts <olly@survex.com> * HACKING: Expand on details of what's required when changing Xapian (discuss documentation requirements, expand on why feature tests are vital). * HACKING: Update section on building debian packages. Sat Apr 08 20:02:19 BST 2006 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.ac: Updated for 0.9.5. Fri Apr 07 23:53:08 BST 2006 Olly Betts <olly@survex.com> * tests/api_anydb.cc,tests/api_db.cc,tests/btreetest.cc: Correct spelling of "existant" to "existent". Fri Apr 07 19:13:24 BST 2006 Olly Betts <olly@survex.com> * configure.ac: We don't use strcasecmp, so don't probe for it. Fri Apr 07 18:30:40 BST 2006 Olly Betts <olly@survex.com> * common/utils.h: Fixes for MSVC7 compilation. Fri Apr 07 17:19:43 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc,backends/quartz/btree.cc: Fixes for MSVC compilation. Fri Apr 07 16:04:08 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_cursor.h: Fix incorrect example code in documentation comment. * backends/flint/flint_table.cc,backends/flint/flint_table.h: Remove unused method FlintTable::find_key(). Fri Apr 07 16:02:40 BST 2006 Olly Betts <olly@survex.com> * AUTHORS: Updated. Fri Apr 07 15:58:31 BST 2006 Olly Betts <olly@survex.com> * debian/control.in: copydatabase and xapian-compact are packaged in xapian-tools, so add them to the documented list of tools included. Fri Apr 07 15:12:43 BST 2006 Olly Betts <olly@survex.com> * net/tcpserver.cc: Set xapian-tcpsrv to allow 5 connections in the listen queue instead of just one. Fri Apr 07 13:28:15 BST 2006 Olly Betts <olly@survex.com> * tests/remotetest.cc: Check mset size in tcpmatch1. Fri Apr 07 01:26:03 BST 2006 Olly Betts <olly@survex.com> * xapian.spec.in: Man pages may be gzipped. Thu Apr 06 14:41:29 BST 2006 Olly Betts <olly@survex.com> * HACKING: aclocal is part of automake, not autoconf. Thu Apr 06 01:29:21 BST 2006 Olly Betts <olly@survex.com> * bin/,examples/copydatabase.cc,examples/delve.cc,examples/quest.cc: In the "--help" output, add "Options:" before the list of options. Thu Apr 06 01:11:31 BST 2006 Olly Betts <olly@survex.com> * xapian-config.in: Tweak to improve help2man output. Wed Apr 05 16:26:15 BST 2006 Fabrice Colin * xapian.spec.in: Package man pages. Wed Apr 05 16:23:49 BST 2006 Olly Betts <olly@survex.com> * Makefile.am,bin/Makefile.am,examples/Makefile.am: Include generated man pages in the distribution tarball. Wed Apr 05 02:48:27 BST 2006 Olly Betts <olly@survex.com> * debian/TODO: Updated. Wed Apr 05 02:44:15 BST 2006 Olly Betts <olly@survex.com> * ./,examples: svn:ignore man pages. Wed Apr 05 02:43:08 BST 2006 Olly Betts <olly@survex.com> * bin/Makefile.am,examples/Makefile.am: No need to make man pages depend on config.h, since the binaries will already. Wed Apr 05 02:42:42 BST 2006 Olly Betts <olly@survex.com> * Makefile.am: Generate man page for xapian-config. Wed Apr 05 00:59:53 BST 2006 Olly Betts <olly@survex.com> * examples/Makefile.am: Use help2man to generate manpages for the installed binaries in examples. Wed Apr 05 00:56:11 BST 2006 Olly Betts <olly@survex.com> * bin/omtcpsrv.cc: Rename to bin/xapian-tcpsrv.cc. * bin/omprogsrv.cc: Rename to bin/xapian-progsrv.cc. * HACKING,bin/Makefile.am,configure.ac: Use help2man to generate manpages for the installed binaries in bin. Tue Apr 04 16:44:54 BST 2006 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Note example of BM25Weight parameters which make set_sort_by_relevance_then_value useful. Tue Apr 04 16:35:58 BST 2006 Olly Betts <olly@survex.com> * api/omenquire.cc,include/xapian/enquire.h,matcher/,tests/api_db.cc: Implement Enquire::set_sort_by_relevance_then_value(). Tue Apr 04 01:05:41 BST 2006 Olly Betts <olly@survex.com> * common/omenquireinternal.h: sort_key no longer needs to be mutable now that MSetSortCmp has been removed. Mon Apr 03 02:03:12 BST 2006 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Removed dead code (class MSetSortCmp). Sun Apr 02 16:37:38 BST 2006 Olly Betts <olly@survex.com> * net/tcpserver.cc: Don't perform a name lookup on the IP address which an incoming connection is from as that could easily slow down the search response - instead just print the IP address itself if output is verbose. Sun Apr 02 13:28:31 BST 2006 Olly Betts <olly@survex.com> * api/omenquire.cc,common/,docs/remote_protocol.html, matcher/multimatch.cc,matcher/networkmatch.cc,matcher/networkmatch.h, net/socketclient.cc,net/socketserver.cc: Change bool sort_by_relevance to enum sort_by in preparation for adding "sort_by_relevance_then_value". Fri Mar 31 22:32:16 BST 2006 Olly Betts <olly@survex.com> * examples/copydatabase.cc,examples/quest.cc: Add --help and --version options. * examples/delve.cc: Tidy up output from --help and --version options. Fri Mar 31 19:01:25 BST 2006 Olly Betts <olly@survex.com> * bin/quartzcheck.cc: Fix SEGV when run with no arguments (introduced by last change). Fri Mar 31 17:57:27 BST 2006 Olly Betts <olly@survex.com> * bin/quartzcheck.cc: Add --version option. Tidy up output from --help. Fri Mar 31 17:43:25 BST 2006 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Include --help and --version in --help output. * bin/quartzdump.cc: Add --help and --version options. Terminate list of long options so that "quartzdump --foo" no longer segfaults. Fri Mar 31 17:27:09 BST 2006 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Tweak --help and --version output for consistency with other binaries. Terminate list of long options so that "quartzcompact --foo" no longer segfaults. Fri Mar 31 16:47:02 BST 2006 Olly Betts <olly@survex.com> * bin/omtcpsrv.cc: Make OPT_HELP and OPT_VERSION positive numbers. Only give synopsis line for --help, not for syntax error. * bin/xapian-compact.cc: List --help and --version in --help output. Terminate list of long options so that "xapian-compact --foo" no longer segfaults. Fri Mar 31 16:20:24 BST 2006 Olly Betts <olly@survex.com> * bin/omtcpsrv.cc: Added --help and --version options. Thu Mar 30 11:51:21 BST 2006 Philip Neustrom * docs/remote_protocol.html: Document messages for requesting and sending a termlist and a document. Wed Mar 29 19:39:05 BST 2006 Olly Betts <olly@survex.com> * backends/flint/flint_termlist.cc,backends/quartz/quartz_termlist.cc: Add missing spaces in debug output. Fri Mar 17 09:22:54 GMT 2006 Olly Betts <olly@survex.com> * api/vectortermlist.h,backends/net/net_termlist.cc, backends/net/net_termlist.h,common/alltermslist.h,common/termlist.h, matcher/branchtermlist.h: Make TermList::positionlist_begin() pure virtual and put dummy implementations in BranchTermList and other subclasses which can't (or don't) implement it. This makes it hard to accidentally fail to implement it in a backend's TermList subclass. * backends/net/net_termlist.h: positionlist_begin() now throws UnimplementedError instead of InvalidOperationError. Fri Mar 17 08:46:52 GMT 2006 Olly Betts <olly@survex.com> * api/omdatabase.cc: There's no need for the MultiTermList wrapper in the common case where we're only dealing with a single database. Fri Mar 17 07:54:54 GMT 2006 Olly Betts <olly@survex.com> * backends/net/net_termlist.h: Remove unused "positions" member. Fri Mar 17 07:07:57 GMT 2006 Olly Betts <olly@survex.com> * backends/multi/multi_termlist.cc,backends/multi/multi_termlist.h, tests/api_posdb.cc: Fix TermIterator::positionlist_begin() to work on TermIterator from Database::termlist_begin(). Fri Mar 17 03:47:04 GMT 2006 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Fix "unused parameter" warning from previous change. Thu Mar 16 05:20:16 GMT 2006 Olly Betts <olly@survex.com> * api/omenquire.cc,include/xapian/enquire.h: Move Enquire::get_matching_terms_end() inline in header. Thu Mar 16 04:28:20 GMT 2006 Olly Betts <olly@survex.com> * api/omdocument.cc,configure.ac,include/xapian/document.h: Move Document::termlist_end() inline in header. Fri Mar 10 04:55:57 GMT 2006 Olly Betts <olly@survex.com> * INSTALL: Improve wording. Fri Mar 10 04:49:09 GMT 2006 Olly Betts <olly@survex.com> * backends/quartz/bcursor.h: Fix incorrect method name in documentation comment. Fri Mar 10 04:39:45 GMT 2006 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Btree::find_key() is unused so remove it. Fri Mar 10 04:29:46 GMT 2006 Olly Betts <olly@survex.com> * HACKING: Note that we now use a lightly patched version of libtool 1.5.22. Fri Mar 10 02:55:48 GMT 2006 Olly Betts <olly@survex.com> * docs/overview.html: Bring up to date. Wed Mar 08 02:32:49 GMT 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix FLAG_BOOLEAN_ANY_CASE to really allow any case combination - previously it only allowed all uppercase or all lowercase. * tests/queryparsertest.cc: Add feature and regression tests for FLAG_BOOLEAN_ANY_CASE. * tests/queryparsertest.cc: Rename test cases to more descriptive names. Tue Mar 07 19:59:54 GMT 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Fix QueryParser's handling of terms with trailing "#", "+", or "-" when set_database has been called and the term doesn't exist in the database with the suffix. Tue Feb 21 21:14:22 GMT 2006 Olly Betts <olly@survex.com> * include/xapian/deprecated.h: Allow xapian-bindings to override deprecation warnings. Tue Feb 21 17:55:17 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc: Note that "flicklock" should be "flintlock"! Tue Feb 21 14:13:08 GMT 2006 Olly Betts <olly@survex.com> * PLATFORMS: Updated. Tue Feb 21 14:12:47 GMT 2006 Olly Betts <olly@survex.com> * HACKING: Update the "how to do a release" list. Tue Feb 21 00:17:40 GMT 2006 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.ac: Updated for 0.9.4. Mon Feb 20 21:15:46 GMT 2006 Olly Betts <olly@survex.com> * api/omdocument.cc,api/omenquire.cc,api/omquery.cc,include/xapian/, queryparser/queryparser.cc: GCC 3.2 actually doesn't like __attribute__((deprecated)) on method definitions, so just put declarations of deprecated methods in the headers, and move definitions into the library. Mon Feb 20 16:04:33 GMT 2006 Olly Betts <olly@survex.com> * tests/api_anydb.cc,tests/api_db.cc,tests/api_nodb.cc: Update uses of deprecated methods and functions. Mon Feb 20 15:47:31 GMT 2006 Olly Betts <olly@survex.com> * include/xapian/document.h: __attribute__((deprecated)) doesn't work on method definitions with default parameters on GCC 3.2 so fix header to overload instead. Mon Feb 20 13:44:14 GMT 2006 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Add documentation comment for Enquire::set_sort_by_value_then_relevance(). Sun Feb 19 23:18:09 GMT 2006 Olly Betts <olly@survex.com> * include/Makefile.am,include/xapian/: Flag deprecated methods such that the compiler gives a warning, for compilers which support such a feature. Sun Feb 19 22:58:55 GMT 2006 Olly Betts <olly@survex.com> * COPYING: Update second occurrence of old FSF address. Sun Feb 19 22:46:49 GMT 2006 Olly Betts <olly@survex.com> * README: Add pointer to HACKING. Change "CVS access" to "SVN access". Sun Feb 19 01:46:00 GMT 2006 Olly Betts <olly@survex.com> * api/version.cc: Correct typo in name of function xapian_revision(). Thu Feb 16 10:23:59 GMT 2006 Olly Betts <olly@survex.com> * xapian-config.in: Oops, fix previous fix (I'd pasted the substituted result from testing the patch...) Thu Feb 16 10:16:31 GMT 2006 Olly Betts <olly@survex.com> * PLATFORMS: Updated from tinderbox. Thu Feb 16 10:15:51 GMT 2006 Olly Betts <olly@survex.com> * xapian-config.in: Need to set exec_prefix and prefix at top of script as they're used by various @SUBSTITUTIONS@. Thu Feb 16 00:09:34 GMT 2006 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.ac: Updated for 0.9.3. Wed Feb 15 21:58:23 GMT 2006 Olly Betts <olly@survex.com> * xapian-config.in: Fix option loop to work on shells other than bash. Wed Feb 15 21:38:38 GMT 2006 Olly Betts <olly@survex.com> * include/xapian/database.h: Tweak wording of a documentation comment. Wed Feb 15 21:34:55 GMT 2006 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Fix documentation comments for the values of QueryParser::feature_flag so doxygen actually pulls out the documentation for them. Add documentation for the parameters of QueryParser::parse_query(). Wed Feb 15 21:26:48 GMT 2006 Olly Betts <olly@survex.com> * bin/quartzcheck.cc: If the database is too broken to open, emit a warning message and bump the error count. Wed Feb 15 21:23:38 GMT 2006 Olly Betts <olly@survex.com> * docs/queryparser.html: Document right-truncation. Wed Feb 15 21:08:37 GMT 2006 Olly Betts <olly@survex.com> * xapian-config.in: Fixed to output usage correctly if no arguments are specified. Mon Feb 13 17:43:08 GMT 2006 Olly Betts <olly@survex.com> * xapian-config.in: Expand dependency_libs recursively. It's not *that* hard, and we're just storing up problems for the future by ignoring the issue. Mon Feb 13 16:57:36 GMT 2006 Olly Betts <olly@survex.com> * HACKING: Devlopers also need makeindex installed for documentation building. Note that dvips and makeindex are usually packaged with TeX. Mon Feb 13 15:16:02 GMT 2006 Olly Betts <olly@survex.com> * configure.ac,xapian-config.in: Some Linux distros have an unhelpful policy of not packaging .la files, and on Linux link_all_deplibs_CXX=no so we don't actually need to link in the dependency_libs. So use the value of link_all_deplibs_CXX from configure to control whether we link against dependency_libs. * xapian-config.in: Factor out common code into shell functions, and tidy up --help output. * xapian-config.in: Re entry "Tue May 11 20:55:56 BST 2004": the (C) dates I mined from CVS included a BrightStation (C), but the few fragments of that version which survive are actually from glib-config (or one of the myriad of *-config scripts which look very like it), and are just standard Bourne shell idioms anyway. Mon Feb 13 13:59:02 GMT 2006 Olly Betts <olly@survex.com> * PLATFORMS: td174 is gone. * PLATFORMS: Compaq C++ 7.1 seems to have better template support (but fails to link binaries). Mon Feb 13 12:22:23 GMT 2006 Olly Betts <olly@survex.com> * HACKING,testsuite/testsuite.cc: XAPIAN_TESTSUITE_PLAIN_OUTPUT -> XAPIAN_TESTSUITE_OUTPUT=plain. Sun Feb 12 18:29:55 GMT 2006 Olly Betts <olly@survex.com> * backends/database.cc,tests/api_nodb.cc: Trying to open a database for reading which doesn't exist now fails with DatabaseOpeningError instead of FeatureUnavailableError. Added regression test nosuchdb1. Thu Feb 09 10:48:10 GMT 2006 Olly Betts <olly@survex.com> * net/socketserver.cc: Add missing '#include <iostream>' when TIMING_PATCH is defined. Wed Feb 08 08:22:09 GMT 2006 Olly Betts <olly@survex.com> * HACKING: Now use libtool 1.5.22 for generating snapshots and releases (includes a number of bug-fixes). * HACKING,docs/doxygen_api_header.html_tmpl, docs/doxygen_full_header.html_tmpl: Now use doxygen 1.4.6 for generating snapshots and releases (nicer output). * docs/doxygen_full_header.html_tmpl: Title "Internal Source Documentation" rather than "Full source documentation". Sun Feb 05 06:19:29 GMT 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: The QueryParser now recognises "AND NOT" as a synonym for "NOT". Mon Jan 16 18:19:26 GMT 2006 Olly Betts <olly@survex.com> * backends/quartz/quartz_utils.h: Fix compiler warning. Mon Jan 16 18:17:27 GMT 2006 Olly Betts <olly@survex.com> * HACKING,backends/flint/,backends/muscat36/,backends/quartz/, bin/quartzcompact.cc,bin/xapian-compact.cc,common/Makefile.am, common/safeerrno.h,net/,tests/btreetest.cc,tests/quartztest.cc, testsuite/backendmanager.cc,testsuite/testsuite.cc: Sorted out a cleaner workaround for Compaq C++'s <errno.h> oddity. Mon Jan 16 17:18:56 GMT 2006 Olly Betts <olly@survex.com> * common/safewindows.h: Fix typo in file description. Mon Jan 16 13:29:48 GMT 2006 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Reverse order of errno includes which seems to make Compaq's C++ compiler happier. Sun Jan 15 23:52:01 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_database.cc,backends/flint/flint_table.cc, backends/quartz/btree.cc,backends/quartz/quartz_database.cc, backends/quartz/quartz_log.cc,configure.ac,net/socketcommon.cc, net/tcpclient.cc,testsuite/backendmanager.cc: Compaq C++ requires sys/errno.h to get ENOENT and most other EXXX codes defined. Sun Jan 15 23:41:15 GMT 2006 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Initialise 2 variables to avoid compiler warnings. Sun Jan 15 23:38:11 GMT 2006 Olly Betts <olly@survex.com> * xapian-config.in: If libxapian.la's dependency_libs contains another .la file, crudely transform to a pair of -L and -l options. Ideally we should recursively pull in dependency_libs from that .la file, but that's too hard to do in sh. Sun Jan 15 21:43:26 GMT 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,queryparser/queryparser.lt: Eliminate ParseAlloc's mallocProc parameter and ParseFree's freeProc parameter - we always pass malloc and free, and get warnings from some compilers because we're passing a C-linkage function for a C++-linkage function pointer. * testsuite/testsuite.cc,testsuite/testsuite.h: Fix functions registered as signal handlers and pass to atexit to have C linkage. Sun Jan 15 14:49:45 GMT 2006 Olly Betts <olly@survex.com> * backends/quartz/quartz_log.cc: #include <string.h> for strerror. Sun Jan 15 04:36:49 GMT 2006 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.cc: Add Stopper::get_description() and SimpleStopper::get_description(). Sat Jan 14 21:43:32 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: Fix backwards logic in snprintf configure test. Also define SNPRINTF_ISO to be an snprintf with ISO C90 semantics for the return value (if one exists), with SNPRINTF being defined to any snprintf which at least performs truncation (which in many cases is sufficient). Sat Jan 14 04:47:33 GMT 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.lt: Fix aCC warnings. Sat Jan 14 04:47:00 GMT 2006 Olly Betts <olly@survex.com> * common/utils.h: Fix aCC warning. Fri Jan 13 18:25:04 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: Turn on more warnings for aCC; suppress existing warning ("Entire translation unit was empty") since it's not useful to us. Fri Jan 13 03:22:26 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: The configure test for snprintf uses memcmp, so we need to "#include <string.h>" for it to work reliably. Wed Jan 11 03:14:30 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: If not cross-compiling, try to actually run a test program built with the C++ compiler, not just link one. Wed Jan 11 03:06:28 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: Note the library version info which 0.9.3 will probably need. Tue Jan 10 22:41:36 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: Fix to actually skip the check for valgrind if VALGRIND is set to an empty value. Tue Jan 10 01:08:10 GMT 2006 Olly Betts <olly@survex.com> * PLATFORMS: Updates from the tinderbox. Mon Jan 09 01:35:57 GMT 2006 Olly Betts <olly@survex.com> * backends/muscat36/io_system.cc,bin/xapian-compact.cc, common/omdebug.cc: More sprintf tweaks. Mon Jan 09 00:58:33 GMT 2006 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Use snprintf if we have it. Sun Jan 08 03:51:52 GMT 2006 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc: Fix conditional compilation of flint backend to use XAPIAN_BUILD_BACKEND_FLINT instead of XAPIAN_BUILD_BACKEND_QUARTZ. Sun Jan 08 02:11:30 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: Disable flint backend by default if building for djgpp or msdos. Sun Jan 08 02:09:48 GMT 2006 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc: Cast NULL to (void*) to avoid "missing sentinel" warning from GCC4. Sat Jan 07 19:09:33 GMT 2006 Olly Betts <olly@survex.com> * HACKING,docs/tests.html: Merge the "running tests" section of docs/tests.html into the similar section in HACKING, and make docs/tests.html refer the reader to HACKING for more information. * HACKING,tests/apitest.cc: Remove OM_TEST_BACKEND. You can now use the "-b" switch to apitest to control which backend is used so it's pretty much redundant. * HACKING,testsuite/testsuite.cc: Rename XAPIAN_SIG_DFL to XAPIAN_TESTSUITE_SIG_DFL. * HACKING,testsuite/testsuite.cc: Add XAPIAN_TESTSUITE_PLAIN_OUTPUT to disable use of ANSI escape sequences in test output. Sat Jan 07 09:13:06 GMT 2006 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Write top-bit set characters using \xXX notation to avoid warnings from Intel's C++ compiler. Sat Jan 07 07:13:25 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: TYPE_SOCKLEN_T fails hard, so only run it if we've successfully run other socket tests. Sat Jan 07 05:29:39 GMT 2006 Olly Betts <olly@survex.com> * queryparser/accentnormalisingitor.h: #include <limits.h> for CHAR_BIT. Fri Jan 06 21:24:01 GMT 2006 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Fix printf type mismatch on 64 bit platforms. Fri Jan 06 18:05:30 GMT 2006 Olly Betts <olly@survex.com> * queryparser/queryparser.cc,queryparser/queryparser.lemony, queryparser/queryparser_internal.h: Replace pair<bool, string> with a simple class BoolAndString - the pair results in a 4328 byte symbol on HP-UX which gets truncated (to 4000 bytes). Fri Jan 06 12:55:37 GMT 2006 Olly Betts <olly@survex.com> * PLATFORMS: Updated from tinderbox. Thu Jan 05 16:12:00 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: Oops, remove hack left over from testing. Thu Jan 05 16:09:06 GMT 2006 Olly Betts <olly@survex.com> * configure.ac: Add sanity check for MS Windows that "find" is Unix-like find, not MSDOS-like. Thu Jan 05 04:27:29 GMT 2006 Olly Betts <olly@survex.com> * INSTALL,PLATFORMS,README: Updated. Mon Dec 19 12:53:38 GMT 2005 Olly Betts <olly@survex.com> * AUTHORS,PLATFORMS: Add success report for Fedora Core 4. Mon Dec 19 12:52:55 GMT 2005 Olly Betts <olly@survex.com> * include/xapian/database.h: Improve a couple of documentation comments. Sun Dec 11 01:33:58 GMT 2005 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Fix handling of "+" terms in a query when the default query operator is AND. * tests/queryparsertest.cc: Add regression test. Fri Dec 09 05:54:33 GMT 2005 Olly Betts <olly@survex.com> * languages/api.cc: Check for malloc and calloc failing to allocate memory and throw an exception. Richard has fix this upstream in snowball, so this is a temporary fix until we import a new version of snowball. Fri Dec 09 02:15:38 GMT 2005 Olly Betts <olly@survex.com> * api/omenquire.cc: Added assertions that the internal ptr isn't NULL to ESet::end() and ESet::get_description(). Fri Dec 09 02:14:13 GMT 2005 Olly Betts <olly@survex.com> * HACKING: Note platforms valgrind now has solid support for; Improve phrasing in a few places. Thu Dec 08 23:13:09 GMT 2005 Olly Betts <olly@survex.com> * INSTALL: Update URL for stlport to sourceforge site, which seems to be where the action is now. Thu Dec 08 04:13:11 GMT 2005 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Fix testsuite harness to work with valgrind on 64 bit platforms (it's not documented, but VALGRIND_COUNT_LEAKS requires the count parameters to be 64 bit on such platforms - i.e. long rather than int as we were using). Tue Nov 01 13:22:50 GMT 2005 Olly Betts <olly@survex.com> * PLATFORMS: Sourceforge have a ppc64 linux box. Xapian builds and works out of the box, so add success report for this platform. Thu Oct 27 09:32:24 BST 2005 Fabrice Colin * xapian.spec.in: Invoke %setup correctly. Sat Oct 01 03:25:50 BST 2005 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Added "--multipass" option to merge postlists in pairs or triples until all are merged. Generally this is faster than an N-way merge, but it does require more disk space for temporary files so it's not the default. Fri Sep 30 18:02:32 BST 2005 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Give pointer to replacements for the deprecated Enquire sorting methods. Wed Sep 28 02:16:03 BST 2005 Olly Betts <olly@survex.com> * bin/xapian-compact.cc: Use a vector<string> to store the list of source databases, in preparation for merging multiple postlist tables in more than one pass. Mon Sep 26 22:22:24 BST 2005 Olly Betts <olly@survex.com> * .: svn:ignore docsource.mk. Mon Sep 26 19:54:15 BST 2005 Olly Betts <olly@survex.com> * bin/: svn:ignore xapian-compact. Mon Sep 26 19:50:13 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Remove trailing whitespace. Wed Sep 21 01:31:23 BST 2005 Olly Betts <olly@survex.com> * api/Makefile.am: Fixed VPATH build which was broken by the addition of version.cc. Tue Sep 20 23:38:05 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc: Don't pass NULL for second parameter of execl(). Tue Sep 20 21:32:04 BST 2005 Olly Betts <olly@survex.com> * api/Makefile.am,api/version.cc,include/xapian/version_h.cc: Add 4 functions to report version information for the library version being used (which may not be the same as that compiled against if shared libraries are in use): xapian_version_string(), xapian_major_version(), xapian_minor_version(), xapian_revision(). Sat Sep 17 14:07:32 BST 2005 Richard Boulton <richard@tartarus.org> * backends/flint/flint_lock.cc: Pass two NULLs to execl() to avoid getting a warning ("not enough variable arguments to fit a sentinel", observed on Ubuntu breezy powerpc). Reported by Sidnei da Silva. Wed Sep 14 23:54:14 BST 2005 Olly Betts <olly@survex.com> * HACKING: Now generate snapshots and releases with automake 1.9.6 (was 1.9.5) and libtool 1.5.20 (was 1.5.18). Tue Aug 30 14:01:55 BST 2005 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.h: Forward declare class InMemoryDatabase to fix compilation with GCC 4.0.1. Fri Aug 19 12:13:39 BST 2005 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc,backends/flint/flint_postlist.cc: Correct uses of termcount which should be doccount. Fri Aug 19 11:48:51 BST 2005 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Fix prototype for ESet::operator[] to take parameter of type termcount instead of doccount. Wed Aug 17 11:50:54 BST 2005 Olly Betts <olly@survex.com> * common/safewindows.h: Fix compilation for cygwin. Tue Aug 16 16:01:36 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_types.h,backends/quartz/quartz_types.h: Using a double to hold the total document length would be a bad idea as we need to store it exactly so kill FIXME comments that suggest perhaps we should. Fri Jul 15 11:11:35 BST 2005 Olly Betts <olly@survex.com> * configure.ac,NEWS,PLATFORMS: Updated for 0.9.2. Fri Jul 15 02:31:11 BST 2005 Olly Betts <olly@survex.com> * preautoreconf: Change directory to the directory that the preautoreconf script is in before doing anything else. Fri Jul 15 01:02:29 BST 2005 Olly Betts <olly@survex.com> * docs/doxygen_api_conf.in,docs/doxygen_full_conf.in,configure.ac: Eliminate TOP_SRCDIR and TOP_BUILDDIR - it's better to just use top_srcdir and top_builddir directly. Fri Jul 15 00:48:39 BST 2005 Olly Betts <olly@survex.com> * configure.ac: Fix superfluous shell quoting. Fri Jul 15 00:28:40 BST 2005 Olly Betts <olly@survex.com> * configure.ac,docs/Makefile.am,preautoreconf,Makefile.am: Generate the list of source files to feed to doxygen by inspecting all the Makefile.am files prior to running autoreconf rather than by using "find" when the user runs ./configure. This speeds up configure, avoids generating docs for random .cc and .h files which aren't part of xapian-core, and avoids problems with picking up FIND.EXE on MS Windows. Thu Jul 14 12:49:36 BST 2005 Olly Betts <olly@survex.com> * common/,matcher/multimatch.cc,matcher/networkmatch.cc, matcher/networkmatch.h,net/socketclient.cc,net/socketserver.cc, tests/api_db.cc: Implement sorting on a value with the remote backend. Wed Jul 13 01:25:17 BST 2005 Olly Betts <olly@survex.com> * include/xapian/database.h,include/xapian/dbfactory.h, net/socketclient.cc: Regularise horizontal whitespace. Tue Jul 05 03:03:57 BST 2005 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Oops, failed to check in header changes for last change. Mon Jul 04 15:50:55 BST 2005 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,queryparser/queryparser.cc: Add flag FLAG_BOOLEAN_ANY_CASE which tells the QueryParser that boolean operators such as "AND", "OR", and "NEAR" should be recognised even if they aren't fully capitalised (so "and", "And", "aNd", etc will work too). Add flag FLAG_WILDCARD which tells the QueryParser to allow right truncation e.g. "xap*". * tests/queryparsertest.cc: Add tests for FLAG_WILDCARD. Sun Jul 03 15:24:12 BST 2005 Olly Betts <olly@survex.com> * backends/database.cc: Fixed to auto-detect database type when opening an existing Flint database as a WritableDatabase. Sat Jul 02 18:49:22 BST 2005 Olly Betts <olly@survex.com> * tests/queryparsertest.cc: Add test case: "-site:xapian.org mail". Sat Jul 02 18:49:03 BST 2005 Olly Betts <olly@survex.com> * common/Makefile.am: Ship safewindows.h header. Sat Jul 02 18:04:18 BST 2005 Olly Betts <olly@survex.com> * bin/quartzcompact.cc,bin/xapian-compact.cc, backends/muscat36/io_system.cc,backends/quartz/btree.cc, backends/quartz/quartz_log.cc,backends/quartz/quartz_database.cc, backends/flint/,common/,net/,tests/,testsuite/backendmanager.cc, testsuite/testsuite.cc: Assorted tweaks towards allowing compilation with MSVC. Sat Jul 02 05:08:54 BST 2005 Olly Betts <olly@survex.com> * backends/quartz/: Always define WIN32_LEAN_AND_MEAN before including windows.h to reduce the amount of stuff it includes and speed up builds; Eliminate Btree::max_key_len as MSVC doesn't like the way we define it and it actually seems simpler to just use BTREE_MAX_KEY_LEN everywhere anyway. * backends/quartz/btree.cc: Added a few more assertions. Sat Jul 02 04:11:01 BST 2005 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony,tests/queryparsertest.cc: Fixed to handle "-site:microsoft.com" where site is a boolean prefix. * queryparser/Makefile.am: Add dependency to generate queryparser_token.h. Thu Jun 30 02:14:16 BST 2005 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Update URL for valgrind FAQ in comment. Thu Jun 30 00:11:52 BST 2005 Olly Betts <olly@survex.com> * include/xapian/enquire.h: More explicit explanation of the "descending docid with boolean weighting" trick for fast date ordered searching. Wed Jun 29 13:16:13 BST 2005 Olly Betts <olly@survex.com> * NEWS: Fix typo: "configurec" -> "configure". Mon Jun 27 04:21:34 BST 2005 Olly Betts <olly@survex.com> * xapian.spec.in: Package xapian-compact. Mon Jun 27 03:41:23 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc,tests/api_wrdb.cc: Fixed mispacking of length for Flint positionlists with more than 127 entries, and added regression test longpositionlist1. Sun Jun 26 02:04:33 BST 2005 Olly Betts <olly@survex.com> * bin/xapian-compact.cc,backends/flint/flint_cursor.h: Let PostlistCursor take ownership of the FlintTable it's iterating over which makes clean-up tidier. * bin/xapian-compact.cc: We were accidentally skipping the first entry in various tables (which essentially meant the first document from each database would go missing when merging databases). Sat Jun 25 23:45:42 BST 2005 Olly Betts <olly@survex.com> * api/omdatabase.cc: Tweak get_lastdocid() code to be a little clearer. Wed Jun 22 20:50:43 BST 2005 Olly Betts <olly@survex.com> * configure.ac,include/xapian/version.h.in,include/xapian/version_h.cc, include/Makefile.am: Eliminate use of "ln -s" when generating include/xapian/version.h since it seems to cause problems on Solaris in some setups and isn't really necessary. Also add dependency mechanism so version.h gets regenerated when the template is changed. Wed Jun 22 18:47:05 BST 2005 Olly Betts <olly@survex.com> * docs/intro_ir.html: Citeseer has moved, so update link. Mon Jun 20 13:33:42 BST 2005 Olly Betts <olly@survex.com> * configure.ac: -Wshadow causes false positives with GCC 3.0.4, so only enable it for 3.1 and up. Mon Jun 20 03:09:07 BST 2005 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: "utilpy" -> "utility"! Fri Jun 17 19:54:44 BST 2005 Olly Betts <olly@survex.com> * COPYING: Update FSF address. Thu Jun 16 18:43:33 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: Fix warnings from older GCC versions. * backends/flint/flint_lock.cc: '#include <signal.h>' so that SIGHUP gets defined reliably. Thu Jun 16 17:54:39 BST 2005 Olly Betts <olly@survex.com> * bin/Makefile.am,bin/xapian-compact.cc: Added new "xapian-compact" program which can compact and merge flint databases in a similar way to how quartzcompact does for quartz databases. Thu Jun 16 01:22:45 BST 2005 Olly Betts <olly@survex.com> * configure.ac: Check for spaces in build directory, source directory, or install prefix and die with a helpful message. Wed Jun 15 01:00:11 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc: It seems we need to explicitly kill the child process. Otherwise when we have two databases locked just closing the connection doesn't cause the child to die. I don't understand why it's needed, but this fix is at least clean. Mon Jun 13 00:13:33 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_alltermslist.cc, backends/flint/flint_alltermslist.h, backends/flint/flint_database.cc: Rewrite of FlintAllTermsList with several fewer member variables. Also fixes a bug (the old version wasn't ignoring the metainfo entry so tests were failing). Sun Jun 12 13:08:16 BST 2005 Olly Betts <olly@survex.com> * backends/quartz/quartz_alltermslist.cc: Disable assertion which is incorrect in a corner case. * tests/api_db.cc: Add test_specialterms2 as a regression test. Sun Jun 12 02:03:52 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Encoding and decoding of position list size, and first and last entries didn't match. Reworked to match using a slightly smaller encoding. * backends/flint/flint_metafile.cc: Bumped format version. Sun Jun 12 02:02:35 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_table.cc: We were failing to append "DB" to the path when opening a table for reading - fixed. Sun Jun 12 02:00:30 BST 2005 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc: When a flint database was requested, we were incorrectly creating a quartz database instead - fixed. Sat Jun 11 17:53:12 BST 2005 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Fix mis-repacking of keys in positionlist table when merging several databases. Thu Jun 09 01:06:35 BST 2005 Olly Betts <olly@survex.com> * Makefile.am,configure.ac: Pass automake options to AM_INIT_AUTOMAKE rather than specifying them in Makefile.am. This way, the version requirements for autoconf and automake are stated close together. Mon Jun 06 19:49:36 BST 2005 Olly Betts <olly@survex.com> * HACKING: Minor updates to release checklist. Mon Jun 06 17:44:19 BST 2005 Olly Betts <olly@survex.com> * NEWS,configure.ac: Updated for 0.9.1. Mon Jun 06 17:28:34 BST 2005 Olly Betts <olly@survex.com> * configure.ac: Describe CC_FOR_BUILD in configure --help output. Mon Jun 06 16:00:26 BST 2005 Olly Betts <olly@survex.com> * PLATFORMS: Updated for 0.9.1. Fri Jun 03 03:49:33 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc,backends/flint/flint_lock.h: Fixed compilation failure on mingw. Fri Jun 03 01:58:08 BST 2005 Olly Betts <olly@survex.com> * PLATFORMS: Updated from tinderbox. Fri Jun 03 00:46:00 BST 2005 Olly Betts <olly@survex.com> * HACKING,NEWS: Snapshot and release tarballs are now bootstrapped using libtool 1.5.18 and automake 1.9.5 (though neither is currently an enforced requirement for bootstrapping). Thu Jun 02 17:36:36 BST 2005 Olly Betts <olly@survex.com> * NEWS: First draft of release notes for 0.9.1. Thu Jun 02 02:05:47 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_record.cc: Fix get_doccount() to not return one too low now that the special entry is no longer in the record table! Wed Jun 01 12:41:59 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Placate Sun's C++ compiler. Tue May 31 19:31:24 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.h, backends/flint/flint_positionlist.cc, backends/flint/flint_metafile.cc: positionlist entries are now stored using interpolative coding (which is significantly more compact). Mon May 30 01:48:17 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_lock.cc,backends/flint/flint_lock.h: Added locking code for MS Windows (untested). Sun May 29 01:00:41 BST 2005 Olly Betts <olly@survex.com> * HACKING,NEWS,testsuite/testsuite.h: Fixed spelling mistakes. Sat May 28 22:51:15 BST 2005 Olly Betts <olly@survex.com> * tests/: svn:ignore .flint/ subdirectory. * tests/Makefile.am: Remove .flint subdirectory on 'make clean'. Sat May 28 22:45:24 BST 2005 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Remove warning that quartz is still in development. Sat May 28 16:24:43 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_lock.h: Removed unused FlintLock method 'operator int()' which was confusing Sun's C++ compiler. Sat May 28 16:18:11 BST 2005 Olly Betts <olly@survex.com> * include/xapian/query.h: Removed superfluous "Query::" which was causing the build to fail with aCC. Sat May 28 14:31:33 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_metafile.cc: Added missing '#include <errno.h>'. Sat May 28 14:15:36 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_positionlist.cc: Simplify code a bit. Sat May 28 02:01:12 BST 2005 Olly Betts <olly@survex.com> * backends/flint/: Move the special item holding the total document length and doc id high water mark from the record table to the postlist table. This means that when appending documents, the insertion point will now always be at the end of the record table. We need to jump around the postlist table to merge anyway. Sat May 28 00:42:25 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_metafile.cc: Change metafile magic to be different from quartz, and make the metafile version a datestamp which we'll change each time the format changes; check the return value of close on the metafile. Fri May 27 22:09:45 BST 2005 Olly Betts <olly@survex.com> * backends/Makefile.am,backends/flint/: Implement new fork+fcntl+exec based locking; Fix new GCC warnings. Fri May 27 22:06:14 BST 2005 Olly Betts <olly@survex.com> * api/omquery.cc: Missed a warning fix. Fri May 27 22:04:18 BST 2005 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h, backends/quartz/quartz_postlist.cc,common/omtime.h,matcher/, net/socketserver.cc,net/socketcommon.cc,net/socketclient.cc, testsuite/btreecheck.h,testsuite/testsuite.cc: Fix GCC warnings from new flags. Fri May 27 20:06:52 BST 2005 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Add missing '#include <config.h>'. Fri May 27 12:20:03 BST 2005 Olly Betts <olly@survex.com> * common/omdebug.h: Oops, change to static_cast reveals that we were discarding const (harmlessly). Fri May 27 12:16:03 BST 2005 Olly Betts <olly@survex.com> * common/omdebug.h: Replace C style cast with static_cast<>. Fri May 27 12:07:48 BST 2005 Olly Betts <olly@survex.com> * configure.ac: -Wendif should be -Wendif-labels. Wed May 25 20:41:39 BST 2005 Olly Betts <olly@survex.com> * configure.ac: Pass more -W flags to g++ (including -Wundef which caught the getopt problem fixed by the previous commit). Wed May 25 20:32:35 BST 2005 Olly Betts <olly@survex.com> * getopt/getopt.cc: Added accidentally pruned #define so that getopt code isn't compiled in when the system uses glibc. Wed May 25 18:36:41 BST 2005 Olly Betts <olly@survex.com> * configure.ac: Add -Wredundant-decls to the default CXXFLAGS for GCC. Wed May 25 03:33:34 BST 2005 Olly Betts <olly@survex.com> * tests/apitest.cc,tests/api_db.cc,tests/api_db.h, testsuite/backendmanager.cc,testsuite/backendmanager.h: apitest now runs tests on flint as well. Wed May 25 03:20:12 BST 2005 Olly Betts <olly@survex.com> * backends/database.cc: When automatically determining which backend to use when creating a WritableDatabase, don't try to open it as both Flint *AND* Quartz (an "else" was missing). Wed May 25 03:19:21 BST 2005 Olly Betts <olly@survex.com> * backends/flint/flint_btreebase.cc: Fix typo in code which prevented a flint database from being opened. Wed May 25 01:10:51 BST 2005 Olly Betts <olly@survex.com> * include/xapian/version.h.in,include/xapian/dbfactory.h, include/xapian/queryparser.h,backends/quartz/,backends/flint/: Flint backend hooked in fully, and everything now builds once more. Tue May 24 02:30:39 BST 2005 Olly Betts <olly@survex.com> * HACKING: Add "email Fabrice" to the release checklist so that RPM spec files don't lag behind. The new svn-tag-release script actually builds the release tarballs rather than just copying snapshots like cvs-tag-release did, so document that. Tue May 24 01:21:18 BST 2005 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Added missing documentation comments. Mon May 23 02:36:06 BST 2005 Olly Betts <olly@survex.com> * configure.ac,backends/database.cc,backends/Makefile.am, backends/flint/: Cloned the quartz backend to form the basis of the new flint backend. Currently it's the same except that the quartz log feature has been removed. When creating a database without a specified backend, quartz is still used unless the environmental variable XAPIAN_PREFER_FLINT is set to a non-empty value. Sat May 21 20:58:48 BST 2005 Olly Betts <olly@survex.com> * xapian.spec.in: Remove bogus %setup line left over from when we packaged xapian-core and xapian-examples together from separate tarballs. Thu May 19 01:12:27 BST 2005 Olly Betts <olly@survex.com> * bin/quartzcheck.cc: Fixed corner case where you couldn't check a single Btree table which was just the DB and baseA/baseB files in a directory (Xapian doesn't produce anything like this, but btreetest does while unit testing the Btree code). Wed May 18 16:59:23 BST 2005 Olly Betts <olly@survex.com> * api/omquery.cc,include/xapian/query.h: Put Query::get_terms_end() inline in header. Wed May 18 16:41:56 BST 2005 Olly Betts <olly@survex.com> * api/omquery.cc,tests/api_nodb.cc: Fix SEGV on get_terms_begin() on an empty Query object. Mon May 16 22:58:09 BST 2005 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Fixed compilation with --enable-debug. Sat May 14 01:02:33 BST 2005 Olly Betts <olly@survex.com> * docs/doxygen_full_conf.in: Fix lingering DOXYGEN_HAVE_DOT reference. Sat May 14 00:36:14 BST 2005 Olly Betts <olly@survex.com> * HACKING: Update information about the SVN tag name to use for debian files. Fri May 13 23:21:54 BST 2005 Olly Betts <olly@survex.com> * HACKING: Updated release instructions to refer to SVN. Fri May 13 20:58:15 BST 2005 Olly Betts <olly@survex.com> * PLATFORMS: Updated for 0.9.0. Fri May 13 02:47:42 BST 2005 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Fix for Sun's C++ compiler. Fri May 13 00:09:07 BST 2005 Olly Betts <olly@survex.com> * NEWS,configure.ac: Updated for 0.9.0. Thu May 12 18:54:11 BST 2005 Olly Betts <olly@survex.com> * docs/doxygen_api_conf.in,docs/Makefile.am: Fix to generate docs for all backend factory functions like 0.8.5 did. Thu May 12 16:59:22 BST 2005 Olly Betts <olly@survex.com> * backends/database.cc,backends/Makefile.am,include/xapian/database.h, include/xapian/dbfactory.h,include/xapian.h,include/Makefile.am: Split off database factory methods into xapian/dbfactory.h - this allows us to fix recent breakage in VPATH builds (caused by the need to include the generated file xapian/version.h from xapian/database.h) without modifying almost every Makefile.am. Wed May 11 16:52:41 BST 2005 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Move prefixed, bracketed subexpression to be an alternative for stop_term which allows "term prefix:(term2 term3)" to parse correctly. Also move non-prefixed, bracketed subexpression to be an alternative for stop_term - this doesn't change how things parse, but makes the grammar simpler. * tests/queryparsertest.cc: Added regression test for the above. * tests/Makefile.am,tests/queryparsertest.cc: Reworked queryparsertest to use the standard testsuite harness. Wed May 11 16:09:32 BST 2005 Olly Betts <olly@survex.com> * include/xapian/query.h: Fix OP_ELITE_SET at 10 (the value it had in 0.8.5). Tue May 10 22:56:45 BST 2005 Olly Betts <olly@survex.com> * docs/Makefile.am: Removed check for DOXYGEN_HAVE_DOT which I removed recently. Tue May 10 01:09:01 BST 2005 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Removed state parameter from TermList::add_term() since it is no longer used. Mon May 09 20:56:40 BST 2005 Olly Betts <olly@survex.com> * docs/Makefile.am: Explicitly set the pool_size for latex, because we seem to overflow the default setting on many systems. * docs/Makefile.am: Use $(MAKE) instead of make. Sun May 08 23:22:37 BST 2005 Olly Betts <olly@survex.com> * configure.ac: Corrected $CC -> $CXX to fix bogus message "Checking for option to enable ANSI C++ mode". * configure.ac: If any tools needed for documentation are missing and we're in maintainer mode, die with a suitable error in configure rather than with strange errors when building the documentation. Sun May 08 01:43:55 BST 2005 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.lemony, queryparser/queryparser_internal.h,queryparser/queryparser.cc: Remove QueryParser::termlist_begin() and termlist_end() since you can just use Query::terms_begin() and terms_end() on the Query object returned by QueryParser::parse_query(). Thu May 05 01:00:26 BST 2005 Olly Betts <olly@survex.com> * common/utils.cc: On WIN32, don't define NOMINMAX if it is already defined. Wed May 04 00:52:04 BST 2005 Olly Betts <olly@survex.com> * matcher/bm25weight.cc: Avoid needing document length if we're simply going to multiply it by zero! Tue May 03 14:07:20 BST 2005 Olly Betts <olly@survex.com> * HACKING: CVS -> SVN; Note that tabs should be 8 spaces. Thu Apr 28 21:13:17 BST 2005 Olly Betts <olly@survex.com> * include/xapian/database.h: Don't declare the backend factory functions if the corresponding backend has been disabled. This means that trying to use a disabled backend will be caught at compile time rather than link time. Thu Apr 28 21:12:43 BST 2005 Olly Betts <olly@survex.com> * docs/intro_ir.html: Corrected two errors. Wed Apr 27 23:04:53 BST 2005 Olly Betts <olly@survex.com> * configure.ac,include/xapian/version.h.in: Define XAPIAN_HAS_xxx_BACKEND for each backend which is enabled. The bindings need this, and user code might find it useful too. Wed Apr 27 03:31:50 BST 2005 Olly Betts <olly@survex.com> * api/omdatabase.cc,api/ompostlistiterator.cc, api/omtermlistiterator.cc,include/xapian/, queryparser/queryparser.cc: Where end iterator is just FooIterator(NULL) put it inline in the header for efficiency. If we ever need to change an implementation, we can easily move methods back into the library and bump the library version suitably. Wed Apr 27 02:40:43 BST 2005 Olly Betts <olly@survex.com> * examples/quest.cc,include/xapian/queryparser.h, queryparser/queryparser.cc,tests/queryparsertest.cc: Rename QueryParser::set_stemming_options() to set_stemming_strategy() - it's a better name and avoids confusion with the old deprecated method called set_stemming_options(). * examples/quest.cc: Added stopword handling. Wed Apr 27 02:39:33 BST 2005 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Fixed documentation comment. Sun Apr 24 03:02:47 BST 2005 Olly Betts <olly@survex.com> * api/omstem.cc: Simplified Xapian::Stem::Internal. Sun Apr 24 02:21:22 BST 2005 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Extended stemlang1 to check that trying to create a stemmer for a non-existent language throws InvalidArgumentError. Fri Apr 22 01:20:57 BST 2005 Olly Betts <olly@survex.com> * include/xapian/document.h: Clearer documentation comments. Thu Apr 21 15:03:42 BST 2005 Olly Betts <olly@survex.com> * docs/overview.html,include/xapian/query.h: Removed lingering references to OP_WEIGHT_CUTOFF. Thu Apr 21 01:46:51 BST 2005 Olly Betts <olly@survex.com> * PLATFORMS: Added success report for Ubuntu 5.04 on x86_64. Mon Apr 18 02:54:10 BST 2005 Olly Betts <olly@survex.com> * docs/stemming.html: Stemming appears to be applicable to Japanese so don't say it isn't! Sat Apr 16 01:08:22 BST 2005 Olly Betts <olly@survex.com> * PLATFORMS: Updated from tinderbox. Fri Apr 15 23:18:58 BST 2005 Olly Betts <olly@survex.com> * languages/: Remove lingering .c versions of snowball generated sources. Fri Apr 15 02:05:27 BST 2005 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser_internal.h, queryparser/queryparser.cc: Implemented QueryParser::set_database() method. * queryparser/queryparser.lemony: Allow searches for C#, etc. If a database has been set, for this and + and - suffixes, check if the term actually exists, and if not, ignore the suffix if the unsuffixed term exists. * tests/queryparsertest.cc: Added tests for # suffix. Thu Apr 14 23:39:57 BST 2005 Olly Betts <olly@survex.com> * include/xapian/stem.h: Rewritten with better documentation comments. * include/xapian/stem.h: Added operator() as preferred alternative to stem_word(). * api/omstem.cc,include/xapian/stem.h: Simplified by restructuring to eliminate a few internal methods. Thu Apr 14 23:35:47 BST 2005 Olly Betts <olly@survex.com> * queryparser/queryparser.cc: Make get_description() debug tracing message category INTRO not API. Thu Apr 14 19:57:06 BST 2005 Olly Betts <olly@survex.com> * include/xapian/queryparser.h,queryparser/queryparser.cc: Added QueryParser::get_description() method (not very descriptive yet!) Thu Apr 14 03:30:09 BST 2005 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Fixed to only include the xapian headers we actually depend on. * include/xapian/queryparser.h: Added \file and \brief documentation comments. Thu Apr 14 02:50:36 BST 2005 Olly Betts <olly@survex.com> * examples/quest.cc,include/xapian.h,tests/queryparsertest.cc: Change xapian.h to automatically include xapian/queryparser.h and fix direct includes of xapian/queryparser.h in tests and examples. Thu Apr 14 01:48:39 BST 2005 Olly Betts <olly@survex.com> * include/xapian/queryparser.h: Added Xapian::SimpleStopper which should be sufficient for the majority of uses. * include/xapian/queryparser.h,queryparser/queryparser_internal.h, queryparser/queryparser.cc: Changed QueryParser::set_stopper() to take a const pointer. Thu Apr 07 15:57:15 BST 2005 Olly Betts <olly@survex.com> * api/omenquire.cc,common/omenquireinternal.h,common/multimatch.h, matcher/multimatch.cc,net/socketserver.cc: Tidy up after sort_bands. Thu Apr 07 15:15:27 BST 2005 Olly Betts <olly@survex.com> * api/omenquire.cc,common/,include/xapian/enquire.h, matcher/networkmatch.cc,matcher/multimatch.cc,matcher/networkmatch.h, net/socketserver.cc,net/socketclient.cc,tests/api_db.cc: Enhanced Enquire sorting API and added ability to reverse sort on a value. * api/omenquire.cc: Enquire::get_description() now includes output of Enquire::Internal::get_description(). Thu Apr 07 13:55:56 BST 2005 Olly Betts <olly@survex.com> * HACKING: Fixed a typo. Thu Apr 07 13:37:02 BST 2005 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h, bin/quartzcompact.cc: The "fuller compaction" code needs changes to btree.h which I'd failed to check in. Also document --fuller in 'quartzcompact --help'. Thu Apr 07 03:35:36 BST 2005 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Added "fuller compaction" mode, which ignores the usual "at least 4 items per block" rule. Thu Apr 07 03:28:32 BST 2005 Olly Betts <olly@survex.com> * PLATFORMS: Added success result for GCC 3.2.3 on FreeBSD 4.8; fixed a typo. Wed Apr 06 23:04:15 BST 2005 Olly Betts <olly@survex.com> * queryparser/queryparser.lemony: Added FIXME comment. Wed Apr 06 22:52:53 BST 2005 Olly Betts <olly@survex.com> * ./: Removed ylwrap from svn:ignore. Tue Apr 05 22:34:16 BST 2005 Olly Betts <olly@survex.com> * queryparser/queryparsertest.cc,queryparser/Makefile.am, tests/Makefile.am: Moved queryparsertest into tests/. Tue Apr 05 22:15:05 BST 2005 Olly Betts <olly@survex.com> * examples/: Updated svn:ignore property. Tue Apr 05 21:34:13 BST 2005 Olly Betts <olly@survex.com> * queryparser/Makefile.am,configure.ac,extra/xapian/queryparser.h, extra/,include/Makefile.am,Makefile.am: Move extra/ to queryparser/; Merge extra/xapian/ into include/xapian/. Tue Apr 05 16:37:32 BST 2005 Olly Betts <olly@survex.com> * m4/xapian.m4: Fix m4 quoting (argh!) Tue Apr 05 16:28:06 BST 2005 Olly Betts <olly@survex.com> * m4/xapian.m4: Fix XO_LIB_XAPIAN to trim off any _svn6789 from XAPIAN_VERSION. Tue Apr 05 16:17:31 BST 2005 Olly Betts <olly@survex.com> * configure.ac: Fix m4 quoting. Tue Apr 05 15:16:38 BST 2005 Olly Betts <olly@survex.com> * configure.ac: #define VERSION in version.h shouldn't include any _svn6789 suffix. Tue Apr 05 13:28:23 BST 2005 Olly Betts <olly@survex.com> * configure.ac: Snapshot generator now appends _svn6789 or similar to the version string, so take that into account when finding REVISION. Tue Apr 05 12:55:52 BST 2005 Olly Betts <olly@survex.com> * ChangeLog.examples,Makefile.am,examples/ChangeLog: Move xapian-examples ChangeLog to ChangeLog.examples in the top-level and include in the tarball. Tue Apr 05 02:18:08 BST 2005 Olly Betts <olly@survex.com> * examples/dir_contents,examples/README: Converted xapian-examples README into a dir_contents file. Tue Apr 05 02:12:31 BST 2005 Olly Betts <olly@survex.com> * docs/overview.html,docs/install.html,configure.ac,Makefile.am, xapian.spec.in,examples/Makefile.am: Moved xapian-examples module to examples subdirectory of xapian-core. Tue Apr 05 00:37:05 BST 2005 Olly Betts <olly@survex.com> * m4/.cvsignore,debian/.cvsignore,matcher/.cvsignore,docs/.cvsignore, tests/.cvsignore,bin/.cvsignore,.cvsignore,autoconf/.cvsignore, extra/.cvsignore,include/xapian/.cvsignore,include/.cvsignore, net/.cvsignore,common/.cvsignore,testsuite/.cvsignore, getopt/.cvsignore,languages/.cvsignore,api/.cvsignore, backends/inmemory/.cvsignore,backends/muscat36/.cvsignore, backends/net/.cvsignore,backends/quartz/.cvsignore, backends/multi/.cvsignore,backends/.cvsignore: Remove .cvsignore files, as they're not used by SVN. Thu Mar 10 01:48:20 GMT 2005 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc,testsuite/index_utils.cc, testsuite/index_utils.h: Rewrite of index_utils code, removing unused and unusual features. * tests/testdata/: Cleaned up data for apitest now that paragraphs don't have the odd minimum number of lines! * tests/testdata/apitest_space.txt: New index_utils code has C-like \ escapes for control characters rather than ^ escapes. Wed Mar 09 22:04:08 GMT 2005 Olly Betts <olly@survex.com> * common/utils.h: Added C_isxdigit and C_isnotxdigit. Wed Mar 09 15:22:35 GMT 2005 Olly Betts <olly@survex.com> * testsuite/index_utils.cc: Fixed ^x to actually decode hex values correctly. Tue Mar 08 16:57:25 GMT 2005 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Minor code tidy. Tue Mar 08 15:48:38 GMT 2005 Olly Betts <olly@survex.com> * api/omenquire.cc,docs/matcherdesign.html,include/xapian/enquire.h, matcher/multimatch.cc,tests/api_db.cc: Removed sort_bands support. Tue Mar 08 04:13:03 GMT 2005 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Fixed to add items for all tables in key order which produces results around 40% more compact (previously in the merge case, this was only happening for the postlist table). Tue Mar 08 03:30:06 GMT 2005 Olly Betts <olly@survex.com> * testsuite/btreecheck.cc: Make default check output fit on a single line (at least for many databases). Tue Mar 08 02:34:41 GMT 2005 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fix warning from newer versions of GCC. Sat Mar 05 00:25:07 GMT 2005 Olly Betts <olly@survex.com> * api/omenquire.cc,common/omenquireinternal.h,include/xapian/enquire.h, matcher/multimatch.cc,tests/api_db.cc: Fixed Xapian::RSet to have the same "it's a handle" copy semantics as most of the other classes. Sat Mar 05 00:23:47 GMT 2005 Olly Betts <olly@survex.com> * testsuite/testsuite.h: Fixed TEST_EQUAL_DOUBLE to use DBL_EPSILON correctly. Fri Mar 04 21:06:38 GMT 2005 Olly Betts <olly@survex.com> * tests/testdata/etext.txt: Stripped carriage returns. Fri Mar 04 21:04:25 GMT 2005 Olly Betts <olly@survex.com> * PLATFORMS: Assorted updates. Thu Mar 03 01:17:07 GMT 2005 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Fixed to allow compacting a single database. Fixed handling of very last term when merging postlists. Wed Mar 02 03:18:30 GMT 2005 Olly Betts <olly@survex.com> * backends/quartz/quartz_values.cc: Values are stored in sorted order so we can stop unpacking the list once we get to one after the one we're looking for. Wed Mar 02 02:59:23 GMT 2005 Olly Betts <olly@survex.com> * bin/quartzcheck.cc: Now checks the structure of all the tables, not just the postlist table, and cross-checks doclen values between termlist and postlist tables. Recognises "--help" option. Should now continue after an error (typically it would crash before), and counts the number of errors found. Now exits with non-zero status if any errors were found. Tue Mar 01 18:48:25 GMT 2005 Olly Betts <olly@survex.com> * xapian.spec.in: '/usr/share' -> '%{_datadir}'. Tue Mar 01 16:21:07 GMT 2005 Olly Betts <olly@survex.com> * tests/api_db.cc: Fixed warning from Sun's C++ compiler. Tue Mar 01 02:02:15 GMT 2005 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Extended to allow merging several quartz databases to produce a single compact quartz database. This allows for faster building - simple index in chunks, then merge the chunks. Tue Mar 01 00:08:12 GMT 2005 Olly Betts <olly@survex.com> * xapian.spec.in: Put the .so in the -devel package (it's only useful for linking to - the .so.* files are all that's needed at runtime). Tue Mar 01 00:05:25 GMT 2005 Olly Betts <olly@survex.com> * Makefile.am,xapian-config.in,xapian.spec.in,extra/.cvsignore, extra/Makefile.am: Eliminated the extra library for the queryparser - it's tiny compared to the main library and having it around just complicates things. Mon Feb 28 23:52:19 GMT 2005 Olly Betts <olly@survex.com> * tests/api_anydb.cc: More set_length() changes. Mon Feb 28 20:31:13 GMT 2005 Olly Betts <olly@survex.com> * net/socketclient.cc,net/socketserver.cc,api/omenquire.cc, api/omquery.cc,api/omqueryinternal.cc,common/, include/xapian/enquire.h,include/xapian/query.h,matcher/: Remove Query::set_length() in favour of an optional length parameter to Enquire::set_query(). Mon Feb 28 15:32:32 GMT 2005 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,common/Makefile.am,common/deleter_vector.h: Removed deleter_vector as it's no longer used anywhere. Mon Feb 28 15:27:42 GMT 2005 Olly Betts <olly@survex.com> * api/omquery.cc,api/omqueryinternal.cc,include/xapian/query.h, matcher/localmatch.cc,tests/api_anydb.cc,tests/api_nodb.cc: Eliminated Query::set_elite_set_size(). Mon Feb 28 15:15:45 GMT 2005 Olly Betts <olly@survex.com> * HACKING: Note how to disable use of VALGRIND on the make check command line, or when using runtest directly. Mon Feb 28 13:36:07 GMT 2005 Olly Betts <olly@survex.com> * common/omqueryinternal.h: Removed out-of-date commented out version of Xapian::Query::Internal. Mon Feb 28 13:07:42 GMT 2005 Olly Betts <olly@survex.com> * backends/quartz/quartz_alltermslist.cc,tests/api_db.cc: Updated (C) dates for changes earlier this year. Mon Feb 28 13:04:41 GMT 2005 Olly Betts <olly@survex.com> * api/omquery.cc,api/omqueryinternal.cc,include/xapian/query.h, matcher/localmatch.cc,matcher/localmatch.h,tests/api_anydb.cc, tests/api_nodb.cc,tests/api_posdb.cc: Eliminated Query::set_window in favour of an optional parameter to the constructor, with the aim of making Query objects immutable once constructed. Mon Feb 28 12:32:26 GMT 2005 Olly Betts <olly@survex.com> * extra/queryparser.lemony: Updated 2 uses of Query::set_window(). Mon Feb 28 04:43:20 GMT 2005 Olly Betts <olly@survex.com> * api/omquery.cc,api/omqueryinternal.cc,common/socketcommon.h, include/xapian/query.h,matcher/,tests/api_anydb.cc: Removed OP_WEIGHT_CUTOFF, since it doesn't actually seem to add useful functionality over using Enquire::set_cutoff(). Mon Feb 28 04:31:58 GMT 2005 Olly Betts <olly@survex.com> * matcher/weightcutoffpostlist.cc: Fix updating of docid in WeightCutoffPostList - the current code works, but can end up doing lots (potentially millions) of pointless loop iterations. Mon Feb 28 02:57:29 GMT 2005 Olly Betts <olly@survex.com> * docs/queryparser.html: Note that + and - work on phrases and bracketed expressions. Fri Feb 25 17:06:23 GMT 2005 Olly Betts <olly@survex.com> * backends/quartz/quartz_alltermslist.cc: Fix skip_to on an allterms TermIterator to set the current term when the skip_to-ed term is in the database. * tests/api_db.cc: Add regression test for this (allterms5). Wed Feb 23 19:07:30 GMT 2005 Olly Betts <olly@survex.com> * configure.ac: Suppress 2 Intel C++ warnings which we can't easily code around, and enable -Werror automatically with --enable-maintainer-mode. * configure.ac: Check that the C++ compiler can actually link a program. AC_LANG_CXX doesn't, and if it can't find a C++ compiler it'll just return "g++" which just leads to a later configure test failing in a confusing way. Tue Feb 22 13:39:07 GMT 2005 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Improved documentation comments. * api/omenquire.cc,include/xapian/enquire.h: MSet::max_size() (which only exists so that MSet is an STL compiler) now returns MSet::size() and is inlined from the header. * include/xapian/enquire.h: Added ESet::max_size() (for STL compatibility). Tue Feb 22 00:18:33 GMT 2005 Olly Betts <olly@survex.com> * configure.ac: corrected "none known for yes" or "none known for no" to "none known for g++-3.2" or similar. * configure.ac: autoconf identifies Intel's C++ compiler as GCC, so probe for which it actually is. Mon Feb 21 21:54:29 GMT 2005 Olly Betts <olly@survex.com> * PLATFORMS: Updated. Mon Feb 21 16:06:32 GMT 2005 Olly Betts <olly@survex.com> * PLATFORMS: Added success report for Intel C++ 8.1 on ia64 Linux (and also GCC 3.2.3 on ia64 Linux). Mon Feb 21 05:33:46 GMT 2005 Olly Betts <olly@survex.com> * backends/multi/: Don't delare methods in headers if they're virtual or large. Mon Feb 21 04:11:23 GMT 2005 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h: Don't delare methods in headers if they're virtual or large. Sun Feb 20 21:13:40 GMT 2005 Olly Betts <olly@survex.com> * common/multimatch.h: Removed unused "class SocketServer;". Sun Feb 20 19:47:01 GMT 2005 Olly Betts <olly@survex.com> * net/socketserver.cc: Fixed typo in debug code. Sat Feb 19 18:44:17 GMT 2005 Olly Betts <olly@survex.com> * PLATFORMS: Added minimal testcase which fails to compile with Compaq's C++ compiler (cxx). Sat Feb 19 04:23:51 GMT 2005 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Change Btree::find_in_block to take a bool indicating if this is a leaf block rather than an offset to subtract when it is. Fri Feb 18 04:27:16 GMT 2005 Olly Betts <olly@survex.com> * HACKING: Added better description of how reference-counted API classes are structured. * include/xapian/postingiterator.h,include/xapian/termiterator.h: More (and better) documentation comments. Fri Feb 18 03:05:40 GMT 2005 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Made full compaction a tiny bit more compact. Improved compaction by a few % in non-full case. Tighter bound on amount of memory to reserve to read the tag into. Fri Feb 18 00:47:41 GMT 2005 Olly Betts <olly@survex.com> * api/omenquire.cc: Removed duplicate "#include <xapian/enquire.h>". Thu Feb 17 01:33:01 GMT 2005 Olly Betts <olly@survex.com> * extra/queryparser_internal.h: Further Sun C++ fixes. Wed Feb 16 05:32:40 GMT 2005 Olly Betts <olly@survex.com> * extra/queryparser.lemony: Attempted fix for Sun's C++ compiler. Tue Feb 15 02:10:35 GMT 2005 Olly Betts <olly@survex.com> * extra/queryparser.cc,extra/queryparser.lemony, extra/queryparser_internal.h: Fixed namespace stuff to keep Sun's C++ compiler happy. Mon Feb 14 21:21:08 GMT 2005 Olly Betts <olly@survex.com> * api/vectortermlist.h: Workaround a shortcoming in Sun's C++ compiler. Mon Feb 14 21:16:05 GMT 2005 Olly Betts <olly@survex.com> * PLATFORMS: Results from upgraded sourceforge x86_64 box. Mon Feb 14 18:36:40 GMT 2005 Olly Betts <olly@survex.com> * extra/xapian/queryparser.h: Fixed compilation error. Mon Feb 14 18:00:12 GMT 2005 Olly Betts <olly@survex.com> * extra/xapian/queryparser.h: Added backward compatibility wrapper for old version of QueryParser::set_stemming_options(). Mon Feb 14 17:50:47 GMT 2005 Olly Betts <olly@survex.com> * extra/xapian/queryparser.h: Added dummy QueryParser::set_database() (currently it ignores the parameter). Mon Feb 14 15:34:13 GMT 2005 Olly Betts <olly@survex.com> * extra/lemon.c: C90, not C99! Mon Feb 14 15:20:11 GMT 2005 Olly Betts <olly@survex.com> * extra/lemon.c: C, not C++. Mon Feb 14 05:49:50 GMT 2005 Olly Betts <olly@survex.com> * extra/: Tweak lemon so we can avoid generating files in builddir in a VPATH build. Mon Feb 14 02:44:17 GMT 2005 Olly Betts <olly@survex.com> * extra/Makefile.am: Update to reflect lempar.c -> queryparser.lt. Mon Feb 14 02:25:53 GMT 2005 Olly Betts <olly@survex.com> * extra/lempar.c,extra/queryparser.lt: Rename lemon template to .lt which is more sensible than calling it lempar.c and should work with VPATH builds. Mon Feb 14 02:12:51 GMT 2005 Olly Betts <olly@survex.com> * extra/Makefile.am: Fixes for VPATH builds. Mon Feb 14 01:43:59 GMT 2005 Olly Betts <olly@survex.com> * configure.ac: Need to AC_SUBST(CC_FOR_BUILD). Mon Feb 14 00:19:45 GMT 2005 Olly Betts <olly@survex.com> * extra/Makefile.am: Added dependency for building queryparser.h. Sun Feb 13 23:30:02 GMT 2005 Olly Betts <olly@survex.com> * extra/queryparser.cc,extra/queryparser_internal.h, extra/xapian/queryparser.h: Fixed to compile with GCC 3.3. Sun Feb 13 23:09:15 GMT 2005 Olly Betts <olly@survex.com> * configure.ac: Enhanced valgrind test to (a) see if --tool=memcheck is needed and (b) see if valgrind actually works (we don't want to try to use an x86 valgrind on an x86_64 box). Mon Jan 17 03:21:29 GMT 2005 Olly Betts <olly@survex.com> * api/vectortermlist.h,extra/,extra/xapian/queryparser.h: Rewritten QueryParser class. Uses Lemon instead of Bison to generate the parser, which enables us to stop using static data, so this class is at last reentrant. It now uses a PIMPL style with reference counted internals like most of the other Xapian classes. And direct access to member variables has gone, which unfortunately forces an API change (bug #39). The rewrite also supports more features than the original did. * HACKING,configure.ac: No longer need Bison. * configure.ac: Need CC_FOR_BUILD to compile Lemon with. Mon Jan 17 02:40:40 GMT 2005 Olly Betts <olly@survex.com> * PLATFORMS: IRIX + SGI C++ now compiles with just two warnings - unused variables in Snowball generated code. Sun Jan 16 03:19:56 GMT 2005 Olly Betts <olly@survex.com> * xapian.spec.in: Don't say "%makeinstall" in a comment since rpm tries to expand it and explodes. Sat Jan 15 03:30:33 GMT 2005 Olly Betts <olly@survex.com> * docs/Makefile.am: Clearer rules for making Postscript doxygen docs. Sat Jan 15 02:45:19 GMT 2005 Olly Betts <olly@survex.com> * HACKING: Note that '#include <limits>' isn't supported by GCC 2.95, and other assorted minor tweaks. Sat Jan 08 16:09:10 GMT 2005 Olly Betts <olly@survex.com> * api/maptermlist.h,backends/inmemory/inmemory_database.h, backends/quartz/quartz_postlist.h,common/,matcher/localmatch.h, matcher/mergepostlist.h,matcher/phrasepostlist.h,net/progclient.cc: Fixes for SGI C++ warnings. Fri Jan 7 13:02:49 GMT 2005 Richard Boulton <richard@tartarus.org> * common/positionlist.h: Correct out-of-date documentation comment. Tue Jan 04 03:05:26 GMT 2005 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Removed unnecessary class declaration. Tue Jan 04 03:02:02 GMT 2005 Olly Betts <olly@survex.com> * configure.ac: Automatically enable ANSI C++ mode for SGI's compiler with '-LANG:std'; check that any automatically determined flags for ANSI C++ mode actually allow us to compile a trivial program - if they don't it probably means the compiler isn't the one we were expecting, but one installed with the same name, so we now drop the flags in this case. Thu Dec 30 00:50:58 GMT 2004 Olly Betts <olly@survex.com> * extra/queryparser.yy: QueryParser::parse_query() was failing to clear termlist and unstem. Fri Dec 24 00:07:28 GMT 2004 Olly Betts <olly@survex.com> * PLATFORMS: Updated from tinderbox. Thu Dec 23 21:13:46 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Fixed GCC compilation warning. Thu Dec 23 17:22:15 GMT 2004 Olly Betts <olly@survex.com> * INSTALL,README: Updated. * README: Don't quote chunks of the GPL - just refer people to the full text in COPYING. Thu Dec 23 16:55:03 GMT 2004 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.ac: Updated for 0.8.5 release. Tue Dec 21 13:26:55 GMT 2004 Olly Betts <olly@survex.com> * docs/quickstart.html: Improved wording. Tue Dec 21 13:03:06 GMT 2004 Olly Betts <olly@survex.com> * HACKING: Updated to reflect extra jobs which cvs-tag-release now does. Mon Dec 20 16:25:21 GMT 2004 Olly Betts <olly@survex.com> * bin/Makefile.am: quartzcompact now uses getopt, so need to link it in our version to build on non-glibc platforms. Mon Dec 20 16:23:28 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Added comment noting why we can't truncate separating keys at higher levels. Thu Dec 16 14:46:28 GMT 2004 Olly Betts <olly@survex.com> * HACKING: Fixed a typo, and improved wording. Mon Dec 13 02:52:52 GMT 2004 Olly Betts <olly@survex.com> * backends/multi/multi_postlist.cc: Updated (C) date. Mon Dec 13 02:50:44 GMT 2004 Olly Betts <olly@survex.com> * languages/: Added missing '#include <config.h>' to .cc files (probably harmless, but it should be included as the first thing any source file does). Mon Dec 13 02:21:28 GMT 2004 Olly Betts <olly@survex.com> * bin/quartzdump.cc: Mark the long options as const. Mon Dec 13 02:20:08 GMT 2004 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Tables sizes will always be a whole number of Kbytes, since the blocksize is, so report the size in K. Also report the change in size as well as the before and after sizes. Mon Dec 13 02:11:59 GMT 2004 Olly Betts <olly@survex.com> * common/netutils.cc: Added missing '#include <config.h>' (probably harmless, but it should be included as the first thing any source file does). Mon Dec 13 01:39:53 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Renamed Btree::compress() to Btree::compact() for consistency with "full_compaction" and "quartzcompact". Also, "compress" is confusing since use "compact" and we use that term in the zlib patch. * backends/quartz/btree.cc: When full_compaction is enabled, don't fill the last few bytes of a block if that would mean we needed an extra item and the overhead for that item would use up more of the next block than we save. This reduces the table size after full compaction by up to 0.2% in my tests! Mon Dec 13 01:39:02 GMT 2004 Olly Betts <olly@survex.com> * backends/multi/multi_postlist.cc: Fixed to build with AssertParanoid enabled. Mon Dec 13 00:41:28 GMT 2004 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Added missing '#include <config.h>' so that largefile support is enabled and we report compression statistics for tables > 2G. * bin/quartzcompact.cc: Added --no-full / -n option to disable full compaction. This may be useful if you want to update the database after compacting it (need to test to see if this option is actually useful). Sun Dec 12 21:26:27 GMT 2004 Olly Betts <olly@survex.com> * xapian-config.in: Previous attempted fix to --libs output was wrong - made it actually work. Sun Dec 12 21:24:44 GMT 2004 Olly Betts <olly@survex.com> * xapian.spec.in: %makeinstall puts the wrong paths in the .la files so use "make DESTDIR=... install" instead. Wed Dec 08 15:59:44 GMT 2004 Olly Betts <olly@survex.com> * NEWS: Bumped the 0.8.4 release date. Wed Dec 08 15:32:46 GMT 2004 Olly Betts <olly@survex.com> * xapian-config.in: Make sure that --libs output doesn't include libxapian.la (from libxapianqueryparser.la's dependencies). Tue Dec 07 18:18:26 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_log.cc: Fixed to compile on mingw. Tue Dec 07 15:57:44 GMT 2004 Olly Betts <olly@survex.com> * NEWS: Updated. * api/omdatabase.cc,backends/database.cc,bin/omtcpsrv.cc, common/database.h,docs/,include/xapian/database.h,tests/api_db.cc: Added constructors to Database and WritableDatabase which fulfil the role that the Auto::open() factory functions currently do. Auto::open() is now deprecated. * api/,backends/inmemory/inmemory_database.cc, backends/quartz/quartz_database.cc,backends/quartz/quartz_postlist.h, common/expandweight.h,common/stats.h,include/xapian.h: #include <xapian.h> no longer pulls in xapian/output.h - this removes the external ability to write a Xapian object to an ostream directly, as it's little used and potentially dangerous ('cout << mset[i];' will compile, but you almost certainly meant 'cout << *mset[i];'). You can get the old effect by writing 'cout << obj->get_description();' instead of 'cout << obj;'. Adjusted all the library sources which relied on xapian/output.h pulling in various other xapian/ headers. * backends/muscat36/da_database.cc,backends/muscat36/db_database.cc: Debug output tweaks. * common/emptypostlist.h,matcher/mergepostlist.cc,matcher/multimatch.cc: Added EmptyPostList::get_maxweight() which always returns 0, so you no longer need to explicitly set a weighting scheme on an EmptyPostList. * common/omdebug.h: Only include omstringstream.h if XAPIAN_DEBUG_VERBOSE is in effect. * extra/Makefile.am: queryparsertest needs to link to libxapian.la explicitly. * net/progclient.cc: Whitespace tweak. * testsuite/: Eliminated barely used sources indexer.h, textfile_indexer.h, and textfile_indexer.cc. * testsuite/backendmanager.cc: Removed unused functions make_strvec() and index_file_to_database(). Wed Dec 01 07:15:46 GMT 2004 Olly Betts <olly@survex.com> * PLATFORMS: SF CF x86_64 machine no longer needs --disable-shared. Tue Nov 30 21:53:33 GMT 2004 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.ac: Updated for 0.8.4 release. Tue Nov 30 03:07:43 GMT 2004 Olly Betts <olly@survex.com> * configure.ac,backends/database.cc,backends/quartz/bcursor.cc, backends/quartz/quartz_database.cc, backends/quartz/quartz_postlist.cc,common/omassert.h, common/omdebug.cc,common/omdebug.h,docs/doxygen_full_conf.in, matcher/localmatch.cc,matcher/multimatch.cc,matcher/stats.cc, tests/apitest.cc,tests/btreetest.cc,testsuite/backendmanager.cc, testsuite/backendmanager.h,testsuite/testsuite.cc: Rename all the MUS_xxx defines to XAPIAN_xxx. Tue Nov 30 02:36:33 GMT 2004 Olly Betts <olly@survex.com> * configure.ac,backends/quartz/btree.cc: Get configure to probe for the prototypes (if any) needed for pread and pwrite. Tue Nov 30 01:09:26 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Instead of trying to coax pread/pwrite prototypes out of the system headers, let's try just providing our own! Mon Nov 29 19:51:13 GMT 2004 Olly Betts <olly@survex.com> * extra/Makefile.am: Moved -no-undefined to the right place, plus we also need to pass @ldflags@ and ../libxapian.la in libxapianqueryparser_la_LIBADD. Mon Nov 29 18:28:57 GMT 2004 Olly Betts <olly@survex.com> * AUTHORS: Updated. Mon Nov 29 18:27:12 GMT 2004 Olly Betts <olly@survex.com> * extra/Makefile.am: Added -no-undefined to libxapianqueryparser_la_LDFLAGS so it builds on mingw. Mon Nov 29 17:46:54 GMT 2004 Olly Betts <olly@survex.com> * tests/Makefile.am: Don't try to run "runtest" as a test - it just needs to be a dependency of the tests. Sat Nov 27 02:58:28 GMT 2004 Olly Betts <olly@survex.com> * backends/inmemory/dir_contents: Trimmed whitespace. Sat Nov 27 02:05:51 GMT 2004 Olly Betts <olly@survex.com> * NEWS: Updated ready for 0.8.4 release. Sat Nov 27 02:03:40 GMT 2004 Olly Betts <olly@survex.com> * tests/Makefile.am: Added a dependency so "make check" regenerates runtest if necessary. Sat Nov 27 02:02:58 GMT 2004 Olly Betts <olly@survex.com> * AUTHORS: Updated. Sat Nov 27 01:59:18 GMT 2004 Olly Betts <olly@survex.com> * README: Minor tweak. Sat Nov 27 01:58:53 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Improved a couple of comments. Sat Nov 27 01:42:32 GMT 2004 Olly Betts <olly@survex.com> * docs/: Trimmed trailing whitespace. Sat Nov 27 01:41:33 GMT 2004 Olly Betts <olly@survex.com> * docs/stemming.html: Reworded text from BrightStation times which talked about "the open source release". Fri Nov 26 17:32:55 GMT 2004 Olly Betts <olly@survex.com> * docs/indexerquickstart.html: Removed dead documentation. Fri Nov 26 17:11:19 GMT 2004 Olly Betts <olly@survex.com> * PLATFORMS: Updated with new cygwin report. Fri Nov 26 17:07:21 GMT 2004 Olly Betts <olly@survex.com> * tests/api_anydb.cc,tests/api_nodb.cc,docs/bm25.html, include/xapian/enquire.h,matcher/bm25weight.cc: Renamed BM25 parameters to match standard naming in papers and elsewhere (A->k3, B->k1, C->k2, D->b), eliminated the extra factor of 2 which our C had, and reordered the parameters to k1, k2, k3. This is an incompatible API change for BM25Weight(), so if you are using custom parameters for BM25 you'll need to update your code. Fri Nov 26 15:20:16 GMT 2004 Olly Betts <olly@survex.com> * tests/runtest.in: Allow VALGRIND environmental variable to override the value we got from configure. Fri Nov 26 03:50:36 GMT 2004 Olly Betts <olly@survex.com> * matcher/bm25weight.cc,matcher/tradweight.cc: termfreq is always exact for matching (we only approximate it for query expansion) so replace code to work around bad approximations with Assert() to make sure this never happens. Fri Nov 26 00:19:22 GMT 2004 Olly Betts <olly@survex.com> * matcher/expandweight.cc,matcher/tradweight.cc: If we estimate the term frequency, ensure it has a sane value (>= r and <= N - R + r) rather than bodging around the problem later on. Thu Nov 25 01:29:36 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Fixed recent cygwin change to actually compile. Thu Nov 25 01:06:03 GMT 2004 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Added --help and --version; Check that the source path and desitination path aren't the same; Report each table name when we start compacting it, and some simple stats on the compaction achieved when we finish. Tue Nov 23 16:19:09 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Fixed m4 quoting problem. Tue Nov 23 12:47:54 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Fix the test for GCC3 used to turn on -Werror with --enable-maintainer-mode. Mon Nov 22 03:08:10 GMT 2004 Olly Betts <olly@survex.com> * PLATFORMS: Assorted updates. Mon Nov 22 02:15:22 GMT 2004 Olly Betts <olly@survex.com> * tests/quartztest.cc: Test with DB_CREATE_OR_OPEN in writelock1. Mon Nov 22 01:37:13 GMT 2004 Olly Betts <olly@survex.com> * common/utils.cc: define NOMINMAX as a cleaner way to prevent windows.h from polluting the namespace and colliding with ANSI C++. Sat Nov 20 14:36:43 GMT 2004 Olly Betts <olly@survex.com> * tests/runtest.in: --logfile-fd was renamed to --log-fd in valgrind 2.1.2 with no support for the old option name, so we must probe to decide which to use. Fri Nov 19 13:18:43 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/bcursor.cc,backends/quartz/bcursor.h: Eliminated Bcursor::get_tag() - it's an internal method only used from one other method, and it now just a trivial wrapper around Btree::read_tag(). Fri Nov 19 04:20:39 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/bcursor.cc,backends/quartz/btree.cc, backends/quartz/btree.h: Factored out near identical code from Btree::find_tag() and Bcursor::get_tag() into Btree::read_tag(). Thu Nov 18 03:27:52 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Applied the Quartz "DANGEROUS" patch, but disabled for now. This way it won't keep being broken by changes to the code. Tue Nov 16 04:29:25 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Ah, it's __sun__/__sun/sun, not __solaris__ (at least for x86 Solaris 9). Tue Nov 16 02:56:44 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Solaris defines __solaris__ not __SOLARIS__. Mon Nov 15 12:40:38 GMT 2004 Olly Betts <olly@survex.com> * tests/api_db.cc: Added new test userweight1 to test user defined matching schemes. Sun Nov 14 05:20:34 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_metafile.cc, backends/quartz/quartz_metafile.h: Removed unused QuartzMetaFile::erase() method. Sun Nov 14 04:55:05 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: For cygwin, use the underlying MoveFile API call for locking, as link() doesn't work on FAT partitions. And don't rely on HAVE_LINK to control whether we use link() otherwise - if the configure test somehow misfires, a compilation error is better than using rename() on Unix as that would cause a second writer to smash the lock of the first. Thu Nov 11 06:58:41 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Removed superfluous statement. Tue Nov 09 23:50:10 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Need to include sys/types.h to define ssize_t, size_t, and off_t before we prototype pread and pwrite for OSF. Tue Nov 09 22:39:00 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Tweaked the pread/pwrite mess to try to get it to work on OSF and x86 Solaris while not breaking it elsewhere. Tue Nov 09 19:29:37 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: More refactoring. Tue Nov 09 17:13:10 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/: More refactoring. Also rearranged the contents of the quartz and btree headers, eliminating btree_types.h in the process. Tue Nov 09 16:47:13 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_metafile.cc: Using fdcloser is less clear in trivial cases. Tue Nov 09 16:38:47 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Close the fd of the lock file before trying to rename it on Windows. Tue Nov 09 07:58:02 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h, backends/quartz/btree_util.h: More refactoring. Tue Nov 09 03:24:59 GMT 2004 Olly Betts <olly@survex.com> * xapian-config.in: Added --swigflags option for use with SWIG. Mon Nov 08 22:54:55 GMT 2004 Olly Betts <olly@survex.com> * configure.ac,xapian-config.in: If flags are needed to select ANSI mode with the current compiler, then make xapian-config --cxxflags include them so that Xapian users don't have to jump through the same hoops we do. Mon Nov 08 04:49:21 GMT 2004 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h,backends/muscat36/, backends/net/net_database.cc,backends/quartz/quartz_database.cc, backends/quartz/quartz_database.h,common/database.h, common/net_database.h,matcher/localmatch.cc: If a database contains no positional information, change NEAR and PHRASE queries into AND queries (as otherwise they'd return no matches at all) (bug#56). * tests/api_wrdb.cc: Added feature test phraseorneartoand1. Mon Nov 08 04:09:04 GMT 2004 Olly Betts <olly@survex.com> * configure.ac,backends/quartz/btree.cc: Improved what we do to turn on pread and pwrite declarations in unistd.h so that it works on OSF and doesn't need the HAVE_GLIBC test, yet still works on OpenBSD. Mon Nov 08 03:55:51 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/,testsuite/btreecheck.cc: More refactoring. Sun Nov 07 20:42:32 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: One more tweak. Sun Nov 07 18:18:19 GMT 2004 Olly Betts <olly@survex.com> * configure.ac,include/xapian/version.h.in: Another tweak to hopefully get version.h generation to work everywhere. Sun Nov 07 15:01:23 GMT 2004 Olly Betts <olly@survex.com> * configure.ac,include/xapian/version.h.in: Terminating lines with 'N' makes Sun's C++ unhappy. Try ',' instead. Sun Nov 07 03:54:42 GMT 2004 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Added test_weight1 which tests the built-in Xapian::Weight subclasses (bug#8). Sun Nov 07 03:36:19 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree_util.h, testsuite/btreecheck.cc: More refactoring. Sun Nov 07 03:33:24 GMT 2004 Olly Betts <olly@survex.com> * matcher/bm25weight.cc,matcher/tradweight.cc: Fixed definitions to match declarations after recent change. Sun Nov 07 02:33:50 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Noted LIBRARY_VERSION_INFO which 0.8.4 will get. Sun Nov 07 02:28:38 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/enquire.h: BoolWeight::unserialise() returns BoolWeight*, etc. BoolWeight::clone() returns BoolWeight *. Sun Nov 07 02:24:17 GMT 2004 Olly Betts <olly@survex.com> * HACKING: Note that C++ style casts are preferable to C style casts; Improve wording in a few places. Sun Nov 07 01:05:58 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree_util.h, testsuite/btreecheck.cc: Moved set_block_given_by() and block_given_by() into Item class. Sat Nov 06 21:16:15 GMT 2004 Olly Betts <olly@survex.com> * configure.ac,include/xapian/version.h.in: Fix generation of version.h to work with aCC -E which concatenates adjacent literal strings. Sat Nov 06 20:58:00 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/,testsuite/btreecheck.cc: Started to refactor the Btree manager by introducing Item and Key classes which take care of handling the on-disk format. Sat Nov 06 15:40:05 GMT 2004 Olly Betts <olly@survex.com> * xapian-config.in,m4/xapian.m4: Pass across ac_top_srcdir and use it if provided to say "configure.ac" or "configure.in" rather than "configure.in (or configure.ac)" in the "Add AC_PROG_LIBTOOL" error message. Sat Nov 06 14:49:05 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Updated comment - we now find SOURCEDOC in 2 goes (as there are no longer any C sources). Sat Nov 06 14:43:16 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Oops, STLPORT_CXXFLAGS *is* used (in xapian-config.in) so revert the change which removes it. Sat Nov 06 13:37:23 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/enquire.h,matcher/bm25weight.cc, matcher/tradweight.cc: Move virtual methods of BM25Weight and TradWeight out of the header. Sat Nov 06 13:19:32 GMT 2004 Olly Betts <olly@survex.com> * configure.ac,include/xapian/version.h.in: Another rework of how include/xapian/version.h is generated - this time to make it work with Sun's C++ compiler again; XAPIAN_VERSION is now a string; Define XAPIAN_REVISION (which is 4 for version 0.8.4). Sat Nov 06 04:22:57 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Don't AC_SUBST(STLPORT_CXXFLAGS) as it is never used that way (it's included in AM_CXXFLAGS). Sat Nov 06 01:03:53 GMT 2004 Olly Betts <olly@survex.com> * testsuite/testsuite.h: Removed incorrect comment about how STRINGIZE is used. Fri Nov 05 14:41:59 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/database.h: Removed a default parameter value from one variant of open_db so that there's only one candidate for open_db(string). Fri Nov 05 14:40:38 GMT 2004 Olly Betts <olly@survex.com> * backends/database.cc: Renamed parameter from keys to values to reflect current external naming. Fri Nov 05 14:38:07 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/version.h.in,configure.ac: Fixed generation of include/xapian/version.h to work with aCC. Thu Nov 04 12:17:50 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Try "-std strict_ansi" for Compaq C++. Thu Nov 04 11:22:52 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Fourth argument to AC_CHECK_HEADERS must be non-empty to make a difference. Thu Nov 04 10:21:25 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Found clean fix for inttypes.h problem (previous bodge didn't work anyway). Thu Nov 04 01:18:55 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Eliminated use of \( \) in sed expression as it appears to cause problems on HP-UX. Thu Nov 04 00:54:11 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Added icky workaround to Compaq C++ oddness - cc can find inttypes.h but cxx can't. Wed Nov 03 23:58:48 GMT 2004 Olly Betts <olly@survex.com> * api/omenquire.cc,common/multimatch.h,matcher/multimatch.cc, net/socketserver.cc: Tweaked to compile with Compaq C++. Wed Nov 03 22:55:04 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Need to put flags to select ANSI C++ mode in CXXFLAGS not AM_CXXFLAGS. Wed Nov 03 22:27:08 GMT 2004 Olly Betts <olly@survex.com> * api/omenquire.cc: Fixed typo. Wed Nov 03 21:58:53 GMT 2004 Olly Betts <olly@survex.com> * api/omenquire.cc: Added explicit cast to try to help Compaq C++ build. Wed Nov 03 21:38:29 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Probe for C++ compiler switches for ANSI mode as early as possible. With Compaq's C++, we need -D__USE_STD_IOSTREAM to successfully #include <streambuf>. Wed Nov 03 21:07:43 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Fixed snprintf configure test. Wed Nov 03 19:37:32 GMT 2004 Olly Betts <olly@survex.com> * common/utils.cc: Fixed typo. Wed Nov 03 19:27:22 GMT 2004 Olly Betts <olly@survex.com> * configure.ac,common/utils.cc: Improved snprintf checking. Wed Nov 03 15:43:54 GMT 2004 Olly Betts <olly@survex.com> * getopt/getopt.cc: Fixed to compile when not using glibc. Wed Nov 03 14:55:29 GMT 2004 Olly Betts <olly@survex.com> * bin/omtcpsrv.cc,bin/quartzdump.cc,common/Makefile.am,common/getopt.h, common/gnu_getopt.h,getopt/,testsuite/testsuite.cc: Major overhaul of getopt use. Move from getopt(), getopt_long(), getopt_long_only() to gnu_getopt(), etc so we don't need to dance around avoiding clashes with getopt() stuff in system headers. The new gnu_getopt.h header is cruft free and hopefully won't trip up assorted compilers or platforms. Wed Nov 03 14:53:39 GMT 2004 Olly Betts <olly@survex.com> * tests/Makefile.am: quartztest doesn't use getopt directly, so no need to link it. Wed Nov 03 12:56:56 GMT 2004 Olly Betts <olly@survex.com> * tests/stemtest.cc: Prune unused #includes. Wed Nov 03 03:47:46 GMT 2004 Olly Betts <olly@survex.com> * getopt/getopt.cc: Defining _NO_PROTO is a really bad idea for C++ code! Wed Nov 03 03:34:06 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Removed AC_HEADER_STDC - Compaq's C++ compiler can't find inttypes.h, but their C compiler can! Wed Nov 03 02:48:04 GMT 2004 Olly Betts <olly@survex.com> * net/tcpclient.cc,net/tcpserver.cc: Use SOCKLEN_T for the type we need to pass to various socket calls, since HPUX defines socklen_t yet wants int in those calls. Wed Nov 03 02:43:26 GMT 2004 Olly Betts <olly@survex.com> * autoconf/type_socklen_t.m4,matcher/networkmatch.cc,net/: If __WIN32__ is defined, we want winsock2.h instead of sys/socket.h. Mingw doesn't seem to even have the latter, so I think previously we've been compiling by picking one up from somewhere random! Wed Nov 03 01:10:01 GMT 2004 Olly Betts <olly@survex.com> * autoconf/type_socklen_t.m4: Rewritten to work with HPUX which helpfully defines socklen_t but doesn't use it! Tue Nov 02 21:29:40 GMT 2004 Olly Betts <olly@survex.com> * getopt/getopt.cc: Fixes for Compaq C++. Tue Nov 02 18:51:43 GMT 2004 Olly Betts <olly@survex.com> * getopt/getopt.cc: Protect getopt definition for possible getopt macro declared in getopt.h. Tue Nov 02 16:52:21 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: AM_CONFIG_HEADER -> AC_CONFIG_HEADERS; removed commented out AC_LANG_SAVE and AC_LANG_RESTORE. Tue Nov 02 15:56:02 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: # comments -> dnl comments. Tue Nov 02 15:54:46 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: $GCC -> $GXX; $CC -> $CXX. Tue Nov 02 15:38:21 GMT 2004 Olly Betts <olly@survex.com> * configure.ac,docs/Makefile.am: Removed SOURCEDOC_C_SRC. Tue Nov 02 06:51:39 GMT 2004 Olly Betts <olly@survex.com> * getopt/getopt.cc,getopt/getopt1.cc: Fixed function declarations to not use K&R C syntax. Tue Nov 02 05:31:57 GMT 2004 Olly Betts <olly@survex.com> * getopt/getopt.cc: Make _getopt_internal extern "C" too. Tue Nov 02 04:43:08 GMT 2004 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,HACKING,PLATFORMS,backends/muscat36/, getopt/Makefile.am,getopt/getopt.cc,getopt/getopt1.cc,languages/: Change the few C sources to be C++. This way we don't need to worry about configure choosing a mismatching pair of compilers, or about whether configure tests with the C compiler don't apply to the C++ compiler, or vice versa. Tue Nov 02 04:31:49 GMT 2004 Olly Betts <olly@survex.com> * backends/muscat36/: More C casts converted to C++ casts. Tue Nov 02 04:28:07 GMT 2004 Olly Betts <olly@survex.com> * matcher/bm25weight.cc,matcher/tradweight.cc: More C casts converted to C++ casts. Tue Nov 02 03:22:52 GMT 2004 Olly Betts <olly@survex.com> * backends/muscat36/: Removed unused test harness sources. Tue Nov 02 02:44:34 GMT 2004 Olly Betts <olly@survex.com> * languages/pool.c,languages/pool.h: Removed unused sources. Tue Nov 02 01:51:30 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: AC_TYPE_SIZE_T causes problems with Compaq C++ when it fails to spot size_t (which is there) and the "#define size_t unsigned long" it adds to config.h breaks "using std::size_t;". Also removed AC_C_CONST as I don't believe that's needed either in a largely C++ library. Both of these have been in configure.in since the very first version so I suspect autoscan decided we wanted them. Mon Nov 01 05:44:02 GMT 2004 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fixed compilation problem on alpha Linux. Mon Nov 01 03:16:36 GMT 2004 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,api/omstem.cc, backends/inmemory/inmemory_database.h, backends/multi/multi_termlist.h,backends/quartz/, extra/queryparser.yy,matcher/,net/tcpserver.cc,tests/api_anydb.cc, tests/api_db.cc,tests/quartztest.cc,testsuite/backendmanager.cc, testsuite/btreecheck.cc,testsuite/testsuite.cc: Changed C style casts to C++ style. The syntax is ugly, but they do make the intent clearer which is a good thing. Mon Nov 01 02:56:31 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Select ANSI iostream implementation for Compaq C++. Mon Nov 01 02:41:58 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Compaq's C++ compiler doesn't know snprintf, yet their C compiler does. Let's try running *all* the configure checks with the C++ compiler, since that's what we compile most code with. Mon Nov 01 01:12:43 GMT 2004 Olly Betts <olly@survex.com> * common/netutils.cc: Fixed to compile. Mon Nov 01 01:06:13 GMT 2004 Olly Betts <olly@survex.com> * common/Makefile.am,common/netutils.cc,common/netutils.h: encode_tname() and decode_tname() aren't really sensible candidates for inlining so move them out of a header. Sun Oct 31 15:34:18 GMT 2004 Olly Betts <olly@survex.com> * configure.ac: Turn on -AA when compiling with HP's aCC. Sun Oct 31 15:10:02 GMT 2004 Olly Betts <olly@survex.com> * testsuite/testutils.cc: Fixed mset_range_is_same() and mset_range_is_same_weights() which were only comparing the first items in the range. Luckily the tests still all pass so this wasn't hiding any bugs. Sat Oct 30 19:15:48 BST 2004 Olly Betts <olly@survex.com> * configure.ac: Disable pread/pwrite on HP-UX as they don't work when LFS in enabled, and we definitely want LFS. Sat Oct 30 16:42:24 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/bcursor.cc: Fix Bcursor::del() which didn't always leave the cursor on the next item like it should. * backends/quartz/quartz_postlist.cc: If we're removing a posting list entirely, often there will only be one chunk, so avoid creating a Bcursor in this case. Sat Oct 30 16:55:19 BST 2004 Olly Betts <olly@survex.com> * languages/header.h: Removed unused #define MAXINT and MININT which were clashing with some header on HP-UX. Sat Oct 30 07:09:33 BST 2004 Olly Betts <olly@survex.com> * docs/bm25.html,docs/intro_ir.html: Reworked to talk about Xapian rather than Muscat. Also improved the appearance of the formulae. Sat Oct 30 06:07:14 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Btree::read_block - debug log the value of p, not the irrelevant contents of the block it points to. Fri Oct 29 22:37:31 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Improved comments. Fri Oct 29 05:10:02 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Fixed ultra-obscure bug in the code which finds a key suitable to discriminating between two blocks in a B-tree branch (discovered by reading the code). Comparing the keys didn't consider the length of the second, so it is possible the code would miscompare. But in reality this is extremely unlikely to happen, and even then would probably just mean that the discriminating key wouldn't be as short as it could be. Fri Oct 29 04:12:09 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Simplified Btree::compare_keys() by removing the last case which was dead code as it was covered by an earlier case. Wed Oct 27 21:17:12 BST 2004 Olly Betts <olly@survex.com> * HACKING,tests/runtest.in: Enhanced runtest to allow it to run test programs under valgrind and other tools (gdb was already supported). * testsuite/testsuite.cc: Point the user to the runtest script if srcdir can't be guessed. And no longer look for the test program in the tests subdirectory of the current directory. * common/omdebug.cc: Removed compatibility code for checking OM_DEBUG_FILE and OM_DEBUG_TYPES. * HACKING: Document that %% in XAPIAN_DEBUG_LOG is substituted with the process-id, and that setting XAPIAN_DEBUG_FLAGS to -1 enables all debug messages. * HACKING: Valgrind now supports x86 FreeBSD and PowerPC Linux. * HACKING: Removed mentions of long-dead configure options --enable-profiling, --enable-purify and --enable-insure. Wed Oct 27 21:16:10 BST 2004 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Document parameters of Enquire::register_match_decider(). Wed Oct 27 21:14:32 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Updated. Wed Oct 13 20:21:38 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Revert the previous change as runtime sized arrays are a g++ extension. Calling new and delete on every call to add_item() is probably unwise. Tue Oct 12 23:40:36 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: split_p is only used by Btree::add_item(), so make it a temporary in that method rather than a class member variable which we need to take care to allocate and deallocate. Mon Oct 11 16:32:10 BST 2004 Olly Betts <olly@survex.com> * tests/btreetest.cc: Fix memory leaks in test_cursor1. Mon Oct 11 02:24:50 BST 2004 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Use 5 tables in the example for how we keep revisions in step, since we use 5 tables in quartz. Thu Oct 07 22:51:28 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: An interrupted update could cause any further updates to fail with "New revision too low" because the new revision was being calculated incorrectly - fixed. Wed Oct 06 15:42:31 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,include/xapian/database.h: Check that any user specified block size is a power of 2. And if the block size passed is invalid, use the default of 8192 rather than throwing an exception. Wed Oct 06 12:19:39 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Updated from tinderbox. Wed Oct 06 01:10:46 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/quartz_document.cc, matcher/multimatch.cc,common/multimatch.h: Fix some warnings from Sun's C++ compiler. Thu Sep 30 22:16:37 BST 2004 Olly Betts <olly@survex.com> * common/utils.cc,common/utils.h: Fixes for win32 and sun's c++ compiler. Thu Sep 30 18:24:20 BST 2004 Olly Betts <olly@survex.com> * common/utils.h,extra/queryparser.yy: Fixed bug which caused misparsing of certain prefixed queries, introduced by C_isXXXXX change. Thu Sep 30 11:09:17 BST 2004 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Modified version of changequery1 fails - the fix is tricky, so just make it SKIP for now. Wed Sep 29 21:33:29 BST 2004 Olly Betts <olly@survex.com> * extra/Makefile.am: Fixed to work when srcdir != builddir. Wed Sep 29 18:52:04 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS,docs/Makefile.am: Workaround odd latex problem. Wed Sep 29 17:10:00 BST 2004 Olly Betts <olly@survex.com> * configure.ac,docs/doxygen_api_conf.in,docs/doxygen_full_conf.in: Updated doxygen conf files for doxygen 1.3.8. Wed Sep 29 16:52:53 BST 2004 Olly Betts <olly@survex.com> * common/utils.cc,common/utils.h,extra/Makefile.am, extra/queryparser.yy,tests/api_anydb.cc,testsuite/index_utils.cc: Provide our own C_isalpha(), etc replacements for isalpha(), etc which always work in the C locale and avoid signed char problems. Tue Sep 28 00:04:11 BST 2004 Olly Betts <olly@survex.com> * common/utils.cc,common/utils.h: rmdir() isn't a sensible candidate for inlining so move it out of the header. Mon Sep 27 17:33:23 BST 2004 Olly Betts <olly@survex.com> * extra/queryparser.yy: Be smarter about when to add a ':' when adding a term prefix. Mon Sep 27 16:01:37 BST 2004 Olly Betts <olly@survex.com> * docs/scalability.html: Added note warning about benchmarking from cold. Mon Sep 27 15:20:13 BST 2004 Olly Betts <olly@survex.com> * HACKING: Note that we use doxygen 1.3.8 for snapshots and releases; Note that --enable-maintainer-mode now automatically enables -Werror with GCC 3.0 or newer. Wed Sep 22 16:49:20 BST 2004 Olly Betts <olly@survex.com> * matcher/: Pruned unneeded #include-s and other tidying, some enabled by the previous change. Wed Sep 22 14:04:27 BST 2004 Olly Betts <olly@survex.com> * common/positionlist.h,matcher/: Moved all of the implementations of the XXXPostList classes from the .h files into the .cc files. All the methods are virtual, so we aren't going to gain anything from being able to inline them. Wed Sep 22 02:58:59 BST 2004 Olly Betts <olly@survex.com> * configure.ac: Automatically add -Werror to CFLAGS and CXXFLAGS if maintainer mode is enabled and we're using GCC3 or newer. Don't do this for older GCCs as GCC 2.95 issues spurious warnings. Wed Sep 22 02:41:41 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_record.cc,backends/quartz/quartz_utils.h: Introduced CASSERT_TYPE_UNSIGNED to replace the common use of CASSERT to assert at compile time that a type is unsigned. Tue Sep 21 15:58:05 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Removed reports from versions prior to 0.7.0. So much has changed that these are of little value. Mon Sep 20 15:24:27 BST 2004 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.ac: Version 0.8.3. Mon Sep 20 15:13:34 BST 2004 Olly Betts <olly@survex.com> * AUTHORS: Updated. Mon Sep 20 14:26:35 BST 2004 Olly Betts <olly@survex.com> * xapian.spec.in: Removed reference to xapian-examples' createdatabase (which is no longer in xapian-examples). Mon Sep 20 13:06:59 BST 2004 Olly Betts <olly@survex.com> * xapian.spec.in: Updated version from Fabrice Colin (incorporating changes from Alan Cox's RPM spec files): split off libs into a separate package to allow 32 and 64 bit versions to be installed concurrently; include binaries from xapian-examples; updated source URLs. Mon Sep 20 03:25:44 BST 2004 Olly Betts <olly@survex.com> * tests/api_anydb.cc: Added regression test for previous bug (test checkatleast1). Mon Sep 20 03:16:14 BST 2004 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fixed segfault with check_at_least when there were no matches. Sun Sep 19 17:54:52 BST 2004 Olly Betts <olly@survex.com> * api/omenquire.cc,common/omdebug.cc: Fixed to compile with debug tracing enabled. Sat Sep 18 19:02:54 BST 2004 Olly Betts <olly@survex.com> * tests/api_db.cc: Updated missed use of omprogsrv to xapian-progsrv. Tue Sep 14 18:09:19 BST 2004 Olly Betts <olly@survex.com> * xapian.spec.in,bin/.cvsignore,bin/Makefile.am,debian/control.in, debian/xapian-tools.install,docs/remote.html,tests/remotetest.cc, testsuite/backendmanager.cc: Rename omtcpsrv to xapian-tcpsrv and omprogsrv to xapian-progsrv. Tue Sep 14 16:25:06 BST 2004 Olly Betts <olly@survex.com> * xapian.spec.in: Fixed mangled URL in last checkin. Tue Sep 14 15:35:00 BST 2004 Olly Betts <olly@survex.com> * xapian.spec.in: Updated URL for tarball. Tue Sep 14 02:49:34 BST 2004 Olly Betts <olly@survex.com> * HACKING: Updated the "how to do a release" tasklist. Mon Sep 13 03:19:47 BST 2004 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.ac: Version 0.8.2. Sat Sep 11 16:39:08 BST 2004 Olly Betts <olly@survex.com> * include/xapian/version.h.in: GCC 3.1 reported the wrong value for __GXX_ABI_VERSION (100 not 101) so check 3.0 and 3.1 by version number, keeping the __GXX_ABI_VERSION check for newer versions. Sat Sep 11 02:57:48 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/bcursor.cc: Fixed and reenabled Bcursor::prev() (not currently used, but it will be useful for running posting lists backwards!) Fri Sep 10 13:13:51 BST 2004 Olly Betts <olly@survex.com> * tests/Makefile.am: Need to ship test data for new test. Thu Sep 09 21:58:37 BST 2004 Olly Betts <olly@survex.com> * tests/api_db.cc: Extended feature test sortrel1 to check interaction with Enquire::set_sort_forward(false). Thu Sep 09 21:46:12 BST 2004 Olly Betts <olly@survex.com> * api/omenquire.cc,common/multimatch.h,common/omenquireinternal.h, include/xapian/enquire.h,matcher/multimatch.cc,net/socketserver.cc, tests/api_db.cc,tests/testdata/apitest_sortrel.txt: You can now specify to sort by value, then relevance, then docid instead of by value then docid. Thu Sep 09 19:30:07 BST 2004 Olly Betts <olly@survex.com> * docs/todo.xml: Removed unused file - todo entries are now in bugzilla. Thu Sep 09 13:11:52 BST 2004 Olly Betts <olly@survex.com> * api/omenquire.cc,common/multimatch.h,common/omenquireinternal.h, include/xapian/enquire.h,matcher/multimatch.cc,net/socketserver.cc: Added optional "checkatleast" parameter to Enquire::get_mset() which allows Omega's MIN_HITS functionality to be implemented in the matcher (a bit more efficient). Thu Sep 09 04:18:32 BST 2004 Olly Betts <olly@survex.com> * NEWS: Updated. Thu Sep 09 03:39:52 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Updated from the tinderbox. Sun's C++ compiler can now build Xapian on sparc! Thu Sep 09 02:19:43 BST 2004 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: Added feature test for WritableDatabase::replace_document() and delete_document() with a unique term (apitest:uniqueterm1. * backends/inmemory/inmemory_alltermslist.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h: Fixed bugs thrown up by the new test. Wed Sep 08 19:45:37 BST 2004 Olly Betts <olly@survex.com> * HACKING: Updated details of Solaris open workaround. Wed Sep 08 19:42:15 BST 2004 Olly Betts <olly@survex.com> * common/utils.h,backends/quartz/btree.cc, backends/quartz/quartz_database.cc,backends/quartz/quartz_log.cc: New version of the Solaris open dance - now we always pass in const char * for the filename, and we don't need the dance on new versions, so only dance if open is defined. Wed Sep 08 16:26:15 BST 2004 Olly Betts <olly@survex.com> * api/omquery.cc,include/xapian/query.h: Renamed Query::is_empty() to Query::empty() for consistency. Keep Query::is_empty() for now as a deprecated alias. Wed Sep 08 16:24:44 BST 2004 Olly Betts <olly@survex.com> * tests/api_wrdb.cc: replace_doc -> replace_doc1; added new test replace_doc2 to test using replace_doc to add a document with a specified docid. Wed Sep 08 16:23:50 BST 2004 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.h: Removed unused member variable "indexing". Wed Sep 08 16:14:47 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: If replace_document() is used to add a document with did greater than lastdocid, raise lastdocid so future calls to add_document() won't clash. Wed Sep 08 16:13:51 BST 2004 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc: Fixed bug Database::replace_document() to work with a document id greater than lastdocid. Wed Sep 08 05:10:01 BST 2004 Olly Betts <olly@survex.com> * common/utils.h: The latest fcntl.h dance causes a compile error on mingw, so don't use it there. Wed Sep 08 03:38:20 BST 2004 Olly Betts <olly@survex.com> * common/utils.h: Use a namespace as an extra step in the Sun fcntl open64 dance. Tue Sep 07 02:33:50 BST 2004 Olly Betts <olly@survex.com> * tests/test.da: Another cruft file removed. Tue Sep 07 02:21:49 BST 2004 Olly Betts <olly@survex.com> * tests/btreetest.cc,tests/quartztest.cc: Moved overwrite1 from quartztest to btreetest. Added const in a few places. Tue Sep 07 02:07:43 BST 2004 Olly Betts <olly@survex.com> * tests/quartztest.cc: Removed overwrite2 test. Digging back in CVS it's been disabled since the day after it added. After all this time it's hard to guess exactly what it was intended to test, so just removing it seems simplest. We already have overwrite1 to test getting DatabaseModifiedError. Tue Sep 07 01:52:25 BST 2004 Olly Betts <olly@survex.com> * tests/btreetest.cc: Cleaned up paths to temporary Btrees. Mon Sep 06 17:51:27 BST 2004 Olly Betts <olly@survex.com> * tests/btreetest.cc,tests/quartztest.cc: Now that QuartzTable, QuartzDiskTable, QuartzBufferedTable, QuartzCursor, QuartzDiskCursor, and QuartzBufferedCursor are gone, move quartztest tests which now just use Btree and Bcursor to btreetest. This is more logical and should help make quartztest less of a monster to compile. Mon Sep 06 12:47:47 BST 2004 Olly Betts <olly@survex.com> * tests/Makefile.am: Ship testdata/apitest_allterms4.txt. Mon Sep 06 02:46:32 BST 2004 Olly Betts <olly@survex.com> * tests/quartztest.cc: Xapian::Database will create the directory for the database so there's no need to create it ourselves. Mon Sep 06 02:40:57 BST 2004 Olly Betts <olly@survex.com> * tests/,tests/testdata/apitest_allterms4.txt: Split off tests which require a writable database backend and tests which should work with any database backend from api_db.cc as it was getting rather large. Fixed simplequery2 to work with backends which don't return the document length (such as the muscat36 backends). Fixed allterms4 to work with muscat36 backends. Mon Sep 06 02:39:29 BST 2004 Olly Betts <olly@survex.com> * backends/muscat36/da_database.cc,backends/muscat36/db_database.cc: Fixed to compile now that internal_end_session() has gone. Mon Sep 06 00:59:36 BST 2004 Olly Betts <olly@survex.com> * tests/runtest.in: Added support for running gdb on a test program, automatically sorting out srcdir and libtool. Mon Sep 06 00:49:11 BST 2004 Olly Betts <olly@survex.com> * tests/apitest_parser.pm: Removed long unused file. Sun Sep 05 22:53:27 BST 2004 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Tweaked documentation comment to stop doxygen parsing a hyphen as a single entry bullet point list. Sun Sep 05 20:20:01 BST 2004 Olly Betts <olly@survex.com> * testsuite/backendmanager.h: Added missing "std::" so code will compile with GCC >= 3. Sun Sep 05 13:54:58 BST 2004 Olly Betts <olly@survex.com> * testsuite/backendmanager.h: Removed superfluous "BackendManager::". Sun Sep 05 01:54:49 BST 2004 Olly Betts <olly@survex.com> * common/utils.cc,common/utils.h: Removed now unused files_exist() function. Sun Sep 05 01:50:35 BST 2004 Olly Betts <olly@survex.com> * tests/api_db.cc,tests/apitest.cc,tests/apitest.h, testsuite/backendmanager.cc,testsuite/backendmanager.h: Cleaned up BackendManager by removing complications only required by absentfile1 test, which can be implemented more directly anyway. Sun Sep 05 01:33:16 BST 2004 Olly Betts <olly@survex.com> * tests/quartztest.cc: Cleaned up rather odd code which is a hangover from when keys and tags weren't simply C++ strings. Sat Sep 04 12:35:12 BST 2004 Olly Betts <olly@survex.com> * HACKING: Updated details of which autotools we require. Fri Sep 03 17:51:59 BST 2004 Olly Betts <olly@survex.com> * configure.ac: Require autoconf 2.59. Noted LIBRARY_VERSION_INFO which 0.8.2 will probably have. * Makefile.am: Require automake 1.8.5. Fri Sep 03 16:40:11 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Updated with results from tinderbox. Fri Sep 03 14:34:00 BST 2004 Olly Betts <olly@survex.com> * autoconf/definedir.m4,autoconf/rjb_find_stlport.m4: Quote macro names to fix warning from newer aclocal. Removed comments about future autodetection of stlport, as the user will always need to decide between the STL supplied with the compiler and stlport. Thu Sep 02 17:55:33 BST 2004 Olly Betts <olly@survex.com> * matcher/andpostlist.cc: Initialise lmax and rmax to 0. Hopefully this will fix SIGFPE on apitest's qterminfo2 on alpha linux. Thu Sep 02 14:50:20 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Not storing the document length and last docid on every add means that the magic key won't always exists when there are records and QuartzWritableDatabase::get_doccount() is sometimes off by one. Fix crudely for now by making sure that the magic key does always exist. Longer term the magic key probably should be in the postlist table but that's an incompatible change. Wed Sep 01 16:15:23 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/: Change QuartzWritableDatabase to store the total document length and the last docid itself rather than tallying added and removed document length and writing the last docid back every time a document is added. This gives cleaner code and a small performance win. Removed XAPIAN_FLUSH_THRESHOLD_LENGTH as we no longer tally the length changes, and made the default flush threshold 10000 documents (was 1000). Wed Sep 01 14:21:19 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Turn on previously commented-out code to make the first key null for blocks more than 1 away from the leaves. It saves disk space for a tiny CPU and RAM cost so is bound to be a win overall. Tue Aug 31 13:55:58 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h, backends/quartz/btree_util.h: Merged Btree::make_index_item() into Btree::enter_key(). Tue Aug 31 01:07:19 BST 2004 Olly Betts <olly@survex.com> * NEWS: Updated in preparation for a release. Tue Aug 31 00:53:35 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Short-cutting Btree::cancel causes problems so disable that for the time being at least. Fri Aug 27 13:09:56 BST 2004 Olly Betts <olly@survex.com> * testsuite/btreecheck.cc: Fixed %% to % (presumably a hangover from converting printf to cout). Thu Aug 26 18:11:50 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Reduce quartz postlist chunk threshold from 2048 to 2000 so that chunks won't get split by the Btree. Thu Aug 26 18:10:27 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_record.cc: Throw DocNotFoundError if we try to delete a record which doesn't exist. Thu Aug 26 17:18:56 BST 2004 Olly Betts <olly@survex.com> * configure.ac: Don't define DATADIR - we no longer use it and clashes with more recent mingw headers. Wed Aug 25 22:47:06 BST 2004 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Started section on quartzcompact. Wed Aug 25 22:45:48 BST 2004 Olly Betts <olly@survex.com> * api/omdatabase.cc,include/xapian/database.h: Changed new WritableDatabase::replace_document(term, doc) method to return the docid which the document was given. Tue Aug 24 13:45:52 BST 2004 Olly Betts <olly@survex.com> * api/omenquire.cc,include/xapian/enquire.h: Added new ESet methods swap(), back() and operator[]. * include/xapian/enquire.h: Added documentation comments for MSet methods size(), empty(), swap(), begin(), end(), back(). * include/xapian/enquire.h: Removed bogus documentation saying that some Enquire methods can throw DatabaseOpeningError. Tue Aug 24 10:24:30 BST 2004 Olly Betts <olly@survex.com> * HACKING: Noted automake 1.8 may be problematic. Tweaked list of release tasks. Tue Aug 24 10:16:38 BST 2004 Olly Betts <olly@survex.com> * matcher/multimatch.cc: If a matchdecider is specified and no matches are requested, the lower bound on the number of matches must be 0 (since the matchdecider could reject all the matches). Mon Aug 23 23:03:12 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Improved the "Db block overwritten" message. The DatabaseCorruptError version now suggests multiple writers may be the cause, while the DatabaseModifiedError version uses less alarming wording and says to call Database::reopen(). Sun Aug 22 14:07:35 BST 2004 Olly Betts <olly@survex.com> * indexer/: Removed the old XML-based indexer framework. It's not worked for ages, and nobody seems interested in ressurecting it. If anyone ever is, they can fetch it back from CVS, but otherwise it's just bulking up CVS checkouts. Sun Aug 22 13:02:25 BST 2004 Olly Betts <olly@survex.com> * Makefile.am,configure.ac,extra/Makefile.am: Moved the older library version information into configure.ac. Sun Aug 22 12:58:21 BST 2004 Olly Betts <olly@survex.com> * HACKING,configure.in,configure.ac,backends/Makefile.am, docs/Makefile.am: Renamed configure.in to configure.ac. Sun Aug 22 12:47:43 BST 2004 Olly Betts <olly@survex.com> * xapian-config.in: Add "(or configure.ac)" in message telling the user to add a line to their configure.in. Sun Aug 22 11:38:55 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree_util.h: Adjusted some Asserts to check c is within blocksize rather than 65536. Added a FIXME for those which can't be trivially changed. Sun Aug 22 10:56:56 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: No need to force a flush on QuartzWritableDatabase::open_document() (the document will read things lazily from the database, and that may trigger a forced flush). * backends/quartz/quartz_database.cc, backends/quartz/quartz_database.h: Eliminated QuartzDatabase::open_post_list_internal() and QuartzDatabase::open_term_list_internal(). Sun Aug 22 01:33:05 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc,backends/quartz/quartz_record.cc, backends/quartz/quartz_record.h: WritableDatabase::get_avlength() no longer forces pending changes to be flushed. This means you can now search a modified WritableDatabase without causing a flush unless the search includes a term whose postlist has pending modifications. Fri Aug 20 20:03:59 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree_util.h,common/,docs/overview.html, docs/quickstart.html,include/xapian/enquire.h, matcher/branchpostlist.h,matcher/multimatch.cc,tests/api_db.cc: Corrected multiple occurrences of "an Xapian::XXX" to "a Xapian::XXX" (presumably these all resulted from replacing "Om" with "Xapian::"). Fri Aug 20 17:38:15 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/,bin/quartzcheck.cc,bin/quartzcompact.cc, bin/quartzdump.cc,tests/btreetest.cc,tests/quartztest.cc: Merged QuartzCursor into Bcursor. Fri Aug 20 13:43:04 BST 2004 Olly Betts <olly@survex.com> * NEWS,backends/quartz/: Fixed the problem with "lazy tag reading" in QuartzCursor and reenable that code. The problem was with deleting the current key, so added QuartzCursor::del() which deletes the current key, leaving the cursor on the next item. Fri Aug 20 13:04:33 BST 2004 Olly Betts <olly@survex.com> * tests/api_db.cc: Reenabled test allterms2, but with the iterator copying parts removed - TermIterator is an input_iterator so that part was invalid. Thu Aug 19 19:19:48 BST 2004 Olly Betts <olly@survex.com> * NEWS: Updated from recent ChangeLog entries. Thu Aug 19 14:09:28 BST 2004 Olly Betts <olly@survex.com> * tests/api_db.cc: Added regression test for bug #37. Thu Aug 19 13:31:37 BST 2004 Olly Betts <olly@survex.com> * matcher/localmatch.cc: Fixed problems handling termweights in queries with the same term repeated (bug #37). Thu Aug 19 13:22:12 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_table.cc: Disable the "lazy tag reading" in QuartzCursor for the time being - it seems to cause problems. Thu Aug 19 12:58:10 BST 2004 Olly Betts <olly@survex.com> * bin/quartzcheck.cc,bin/quartzdump.cc: Added calls to QuartzCursor::read_tag(). Thu Aug 19 12:56:53 BST 2004 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Removed superfluous clear() of a map. Thu Aug 19 12:56:08 BST 2004 Olly Betts <olly@survex.com> * tests/api_db.cc: Corrected a comment - a pure boolean query has all weights set to 0, not 1. Mon Aug 16 15:41:33 BST 2004 Olly Betts <olly@survex.com> * docs/: Removed unused and very out of date class diagrams in dia format. Doxygen generates similar but up-to-date diagrams automatically anyway. Mon Aug 16 15:27:30 BST 2004 Olly Betts <olly@survex.com> * xapian.spec.in,debian/libxapianVERSION-dev.install, extra/.cvsignore,extra/Makefile.am,extra/omparsequery.h, include/Makefile.am,include/om/.cvsignore,include/om/dir_contents, include/om/om.h,m4/xapian.m4: Removed the compatibility layer which allowed programs written against the pre-0.7.0 API to be compiled. Mon Aug 16 15:08:08 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_table.cc: Fixed QuartzCursor::find() to work again after the last change. Mon Aug 16 14:43:08 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/,tests/quartztest.cc: QuartzCursor no longer automatically reads the tag - you have to call read_tag() to get it read. This speeds up iterator over all the terms in a database. Also commented out QuartzCursor::prev as it's unused and untested (at least in its latest form). Mon Aug 16 12:39:56 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree_types.h: Made Bcursor.rewrite bool rather than int. Mon Aug 16 12:35:07 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Calculate and store the latest revision number, rather than storing the other one and working out which is newer every time we're asked. Mon Aug 16 12:32:59 BST 2004 Olly Betts <olly@survex.com> * docs/install.html: We haven't "only [...] UNIX" for ages, so don't claim we are. We use libtool as well as autoconf and automake. Link to the CVS snapshots. And also mention downloading omega. Mon Aug 16 12:20:04 BST 2004 Olly Betts <olly@survex.com> * docs/index.html: exaplains -> explains. Mon Aug 16 12:18:34 BST 2004 Olly Betts <olly@survex.com> * docs/overview.html: Removed references to "our company" (meaning BrightStation) and to the now defunct special parameters which the inmemory backend accepted to cause deliberate errors for testing. Sun Aug 15 23:48:20 BST 2004 Olly Betts <olly@survex.com> * tests/btreetest.cc: item_count -> get_entry_count(); revision_number -> get_open_revision_number(). Sun Aug 15 23:15:34 BST 2004 Olly Betts <olly@survex.com> * backends/Makefile.am,backends/inmemory/Makefile.am: Fixed to compile with --disable-inmemory (bug #33). Sat Aug 14 18:56:06 BST 2004 Olly Betts <olly@survex.com> * Makefile.am: Improved library versioning comment. Sat Aug 14 18:55:17 BST 2004 Olly Betts <olly@survex.com> * docs/overview.html: Fixed om_queryop to Xapian::Query::op, and added missing OP_* codes to the list. Sat Aug 14 17:44:39 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/bcursor.h,backends/quartz/btree.cc, backends/quartz/btree.h,bin/quartzcompact.cc,docs/quartzdesign.html: Updated quartz design docs to reflect recent changes. Also pulled out the Btree and Bcursor API docs and slotted them in as doxygen documentation comments - this way they're much more likely to be kept up-to-date. Sat Aug 14 15:59:07 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Removed unused Btree default ctor. Sat Aug 14 15:45:28 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Don't redundantly store next_revision - it's always just revision_number + 1. Sat Aug 14 15:30:30 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Implemented Btree::cancel directly rather than closing and reopening the Btree. Sat Aug 14 14:46:43 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/,bin/quartzcheck.cc,bin/quartzcompact.cc, bin/quartzdump.cc,tests/btreetest.cc,tests/quartztest.cc, testsuite/btreecheck.cc,testsuite/btreecheck.h: Eliminated QuartzTable which had become just a thin wrapper around Btree. Fri Aug 13 19:19:02 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: QuartzPostList::move_to_chunk_containing now calls next_chunk if required, rather than forcing the caller to do the fix-up. Fri Aug 13 17:54:53 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc, backends/quartz/quartz_database.h: Added tunable flush thresholds - set XAPIAN_FLUSH_THRESHOLD=5000 to flush every 5000 documents or XAPIAN_FLUSH_THRESHOLD_LENGTH=1000000 to flush every 1000000 total change in document length. Set both to flush whichever is reached first. Set neither and the default is to flush every 1000 documents as before. Fri Aug 13 15:54:21 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/: Removed no-longer-used Btree::erase and QuartzTable::erase methods. Fri Aug 13 15:37:15 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/quartz_database.cc: Changed Btree::create() to remove any pre-existing alternate base file, so we no longer need to call Btree::erase before Btree::create. Fri Aug 13 15:37:15 BST 2004 Olly Betts <olly@survex.com> * NEWS: Updated with changes since last release. Fri Aug 13 15:36:04 BST 2004 Olly Betts <olly@survex.com> * tests/api_posdb.cc,tests/quartztest.cc: Reworked quartztest's positionlist1 into a generic api test as apitest's poslist3. Thu Aug 12 16:26:42 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_table.cc,backends/quartz/quartz_table.h, bin/quartzcheck.cc,bin/quartzdump.cc: QuartzCursor now has a Bcursor member, rather than an AutoPtr<Bcursor> member. Thu Aug 12 14:13:37 BST 2004 Olly Betts <olly@survex.com> * tests/btreetest.cc: Fixed up in line with the recent refactoring. Thu Aug 12 13:23:31 BST 2004 Olly Betts <olly@survex.com> * m4/xapian.m4: XO_LIB_XAPIAN now AC_SUBSTs XAPIAN_VERSION. Thu Aug 12 13:21:36 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/,tests/quartztest.cc: Refactored, replacing Quartz*Manager with Quartz*Table. Thu Aug 12 02:00:58 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/: Merged QuartzTableManager into QuartzDatabase. Wed Aug 11 23:40:34 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc,backends/quartz/quartz_database.h: Eliminated buffered_tables member of QuartzWritableDatabase. Wed Aug 11 21:26:35 BST 2004 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc, backends/quartz/quartz_database.cc,backends/quartz/quartz_database.h, docs/todo.xml: WritableDatabase::replace_document can now be used to add a document with a specific docid (to allow keeping docids in sync with numeric UIDs from another system). Wed Aug 11 20:09:15 BST 2004 Olly Betts <olly@survex.com> * api/omdatabase.cc,include/xapian/database.h: Added replace_document and delete_document variants which take a unique id term name rather than a document id. * include/xapian/database.h: Better documentation for replace_document and delete_document. Wed Aug 11 16:15:10 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/,bin/quartzcheck.cc,bin/quartzdump.cc, tests/quartztest.cc: Eliminated QuartzBufferedTable. Wed Jun 30 20:34:08 BST 2004 Olly Betts <olly@survex.com> * NEWS: Fixed an unwrapped line. Wed Jun 30 20:32:31 BST 2004 Olly Betts <olly@survex.com> * HACKING: Added note about the cvs-tag-release script. Wed Jun 30 19:05:45 BST 2004 Olly Betts <olly@survex.com> * HACKING,NEWS,PLATFORMS,configure.in: Version 0.8.1. Wed Jun 30 14:23:20 BST 2004 Olly Betts <olly@survex.com> * AUTHORS,PLATFORMS: Updated. Tue Jun 29 23:24:59 BST 2004 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Make emptyquery1 check that Query("") causes an InvalidArgumentError exception. Tue Jun 29 17:29:03 BST 2004 Richard Boulton <richard@tartarus.org> * Makefile.am: Remove Debian files from distribution tarballs, since there will often be multiple patch releases for each release. Debian files will be available from an apt repository in future. Mon Jun 28 01:29:00 BST 2004 Olly Betts <olly@survex.com> * NEWS: Mostly updated for 0.8.1 release. Sun Jun 27 23:37:01 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Fixed bug in postlist merging. Sat Jun 26 00:51:04 BST 2004 Olly Betts <olly@survex.com> * AUTHORS: Add Malcolm Baldridge (helped fix a problem with building the PHP bindings with newer versions of SWIG and PHP). Fri Jun 26 00:29:10 BST 2004 Olly Betts <olly@survex.com> * HACKING,api/ompositionlistiterator.cc,api/ompostlistiterator.cc, api/omtermlistiterator.cc,include/xapian/: MSetIterator and ESetIterator are now bidirectional iterators (rather than just input iterators); Fixed post-increment forms of PostingIterator, TermIterator, PositionIterator, and ValueIterator so that *i++ works (as it must for them to be true input iterators). Thu Jun 24 18:03:46 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Added success report for Slackware Linux 9.1. Mon Jun 21 16:33:16 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Corrected -> to . so code compiles with debug enabled. Mon Jun 21 03:25:24 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Eliminated two calls to abort() - throw exceptions instead. Mon Jun 21 03:21:14 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Finish backing out incorrect change from "Fri May 07 03:16:29 BST 2004" - failed to change two lines back before. Fri Jun 18 16:48:10 BST 2004 Richard Boulton <richard@tartarus.org> * configure.in: Fix typo (STLPORT_CXXLAGS -> STLPORT_CXXFLAGS) Thu Jun 17 03:46:32 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Cleaned up code to track add vs delete vs modify of a posting list entry since we now force a flush if an entry is about to be retouched. Thu Jun 17 03:44:21 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Back out incorrect change from "Fri May 07 03:16:29 BST 2004". We do need to call get_or_make_tag in this case because we're modifying the tag. Wed Jun 16 15:40:21 BST 2004 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.h: Implemented get_lastdocid() for InMemory backend. Wed Jun 16 02:39:41 BST 2004 Olly Betts <olly@survex.com> * configure.in: Note the value of LIBRARY_VERSION_INFO which 0.8.1 would get if it were released now. Wed Jun 16 02:39:11 BST 2004 Robert Pollak <robert.pollak@fabasoft.com> * api/omdatabase.cc, include/xapian/database.h, backends/database.cc, common/database.h, backends/quartz/quartz_database.cc, backends/quartz/quartz_database.h, backends/quartz/quartz_record.cc, backends/quartz/quartz_record.h: New method Database::get_lastdocid for re-synchronizing an old quartz index. Wed Jun 16 02:05:07 BST 2004 Richard Boulton <richard@tartarus.org> * matcher/multimatch.cc: When collapsing, keep track of the number of collapses performed, and use this information to modify the bounds and estimate of the number of matches. * tests/api_db.cc: Added tests for this. * include/xapian/enquire.h: Update documentation comments for MSet::get_matches_*() functions to make clear that collapsing and cutoffs are taken into account. (Previously, the most likely interpretation of the comments was that they wouldn't be taken into account, but the implementation was that percentage cutoffs were taken into account.) Due to this ambiguity, I think it is reasonable to say this isn't an API change. Wed Jun 16 01:55:29 BST 2004 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Rearranged code so that the behaviour can be easily seen to be unchanged, but so that it's closer to the result of applying Richard's patch for bug #31. Wed Jun 16 01:27:09 BST 2004 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Trimmed extra whitespace. Tue Jun 15 15:40:11 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Fixed PostlistChunkReader to take a copy of the postlist data being read to avoid problems with reading data from a string that's been deleted. Tue Jun 15 15:26:54 BST 2004 Olly Betts <olly@survex.com> * HACKING: Updated the list of tasks required for a new release. Tue Jun 15 15:24:04 BST 2004 Olly Betts <olly@survex.com> * Makefile.am,configure.in,extra/Makefile.am: Unify the shlib version numbers (the small benefit of tracking them individually makes it hard to justify the extra work required, and having one version simplifies debian packaging too). Tue Jun 15 14:52:36 BST 2004 Robert Pollak <robert.pollak@fabasoft.com> * extra/xapian/queryparser.h: Fixed memory leaked upon QueryParser destruction. Fri Jun 11 02:18:35 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Refactored a loop. Fri Jun 11 02:17:05 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Fixed bug which meant we sometimes failed to remove a posting when deleting or replacing a document. Fri Jun 11 02:16:16 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc, backends/quartz/quartz_postlist.h: Merged move_to() into skip_to(). Fri Jun 11 02:14:56 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Fixed typo in comment. Thu May 27 15:41:45 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Eliminated the split cursor - we only actually need a single block buffer to handle splitting blocks. Wed May 26 04:02:18 BST 2004 Olly Betts <olly@survex.com> * include/om/om.h,include/xapian/errortypes.h: Removed several unused Xapian::Error subclasses (these were used by the indexer framework which we decided was a failed experiment). Wed May 26 01:54:13 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: More DEBUGCALL tracing added; split_root now uses level member rather than a parameter. Sun May 23 00:56:41 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Merge split_off() into add_item(). Sat May 22 01:28:58 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Another DEBUGCALL. * backends/quartz/quartz_table.h: Improved comments. Fri May 21 23:17:01 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Removed unnecessary assignment which is a hangover from the shared level code we removed long ago. Fri May 21 20:31:56 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/,docs/quartzdesign.html,testsuite/btreecheck.cc: Removed overwritten flag (unused as we throw an exception anyway). If Btree is writable, throw DatabaseCorruptError if we detect overwritten. Make use of bool return types consistent. Removed documentation of Btree error codes (the error codes themselves were removed a month ago). Thu May 20 03:50:46 BST 2004 Olly Betts <olly@survex.com> * include/xapian/database.h: Remove references to sessions in doxygen comments. Thu May 20 03:44:28 BST 2004 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/database.cc, backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/muscat36/da_database.h,backends/muscat36/db_database.h, backends/net/net_database.cc,backends/quartz/quartz_database.cc, backends/quartz/quartz_database.h,common/database.h, common/net_database.h: Clean up the backend interface. Thu May 20 03:18:30 BST 2004 Olly Betts <olly@survex.com> * backends/database.cc,backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h,backends/muscat36/, backends/net/net_database.cc,backends/quartz/quartz_database.cc, backends/quartz/quartz_database.h,common/database.h: Stripped out the session machinery - all that is actually required is to ensure that any unflushed changes are flushed when the dtor runs. Mon May 17 01:04:17 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h, backends/quartz/btree_types.h: Reworked split_p and split_n members of Cursor into a separate C_split cursor. This reduces the memory overhead of each Bcursor (and hence each QuartzPostList). Sat May 15 01:29:40 BST 2004 Olly Betts <olly@survex.com> * docs/quickstart.html: Corrected lingering reference to "om.h" and note that we need <iostream>. * docs/quickstartindex.cc.html,docs/quickstartexpand.cc.html, docs/quickstartsearch.cc.html: Add <iostream>. * AUTHORS: Add John Ward for pointing out the above problems. Fri May 14 00:59:51 BST 2004 Olly Betts <olly@survex.com> * HACKING: Added the start of a list of subtasks when doing a release. Currently it's always me that does this, but it may not always be and anyhow it'll help me to have a list to run through. Fri May 14 00:59:21 BST 2004 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Thu May 13 17:14:18 BST 2004 Olly Betts <olly@survex.com> * extra/queryparser.yy: When stripping non-alphanums prior to reparsing, keep dots ('.'). * extra/queryparsertest.cc: Pruned near-duplicate queryparsertest testcases. Thu May 13 12:09:19 BST 2004 Olly Betts <olly@survex.com> * docs/quartzdesign.html: "interger" -> "integer". Thu May 13 11:20:44 BST 2004 Olly Betts <olly@survex.com> * Makefile.am: Removed bogus extra line added by last change. Thu May 13 11:11:31 BST 2004 Olly Betts <olly@survex.com> * configure.in,Makefile.am,autoconf/.cvsignore,autoconf/Makefile.am, m4/.cvsignore,m4/Makefile.am: Removed trivial m4/Makefile.am and and autoconf/Makefile.am and do the work from the top level Makefile.am instead. It's easy to see the structure this way, and it also removes a couple of recursive make invocations. Thu May 13 09:32:22 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Added DEBUGCALL tracing; Stripped out C_ parameters where they are always the internal cursor C. Thu May 13 00:05:45 BST 2004 Olly Betts <olly@survex.com> * extra/queryparser.yy,extra/queryparsertest.cc: If we fail to parse a query, try stripping out non-alphanumerics and reparsing. Wed May 12 14:32:54 BST 2004 Olly Betts <olly@survex.com> * common/omtime.h: Corrected file description. Wed May 12 02:08:44 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Fixed typos in comment. Tue May 11 20:55:56 BST 2004 Olly Betts <olly@survex.com> * xapian-config.in: Added comment saying why we filter out -I/usr/include; Removed no longer used "optarg" stuff; Added licence boilerplate text (with (C) dates mined from CVS). Tue May 11 13:42:38 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/bcursor.cc,backends/quartz/btree.cc: Added lots of assertions. Tue May 11 09:29:51 BST 2004 Richard Boulton <richard@tartarus.org> * debian/: Added basic Debian packaging. Mon May 10 17:17:45 BST 2004 Olly Betts <olly@survex.com> * docs/quartzdsign.html: Corrected various pieces of out of date information. Mon May 10 01:30:35 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.h: Removed unused forward declaration of class BtreeCheck. Mon May 10 01:30:24 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Fixed typo. Sat May 08 20:14:40 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/: Pruned some #include-s which were unused, or could be replaced by forward class declarations. Fri May 07 14:38:10 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Slight efficiency tweak to the "deleting the whole posting list" case. Fri May 07 03:16:29 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Removed another unnecessary use of QuartzBufferedTable::get_or_make_tag(). Fri May 07 02:56:22 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Initialise prev_ptr and next_ptr at open time for a writable Btree. Fri May 07 02:46:53 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Remove unneeded call to QuartzBufferedTable::get_or_make_tag() in a case when we're using a cursor which has already fetched the tag. Thu May 06 23:30:12 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Improved a comment. Thu May 06 22:20:43 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Added a few more assertions. Thu May 06 02:25:33 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_positionlist.cc, backends/quartz/quartz_utils.h: Added SON_OF_QUARTZ define to disable incompatible changes to database formats by default, and use it to control the docid encoding for keys such that we're always inserting at the end of the table when added new documents. Thu May 06 02:23:02 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Improved a FIXME comment. Thu May 06 00:32:17 BST 2004 Olly Betts <olly@survex.com> * extra/queryparsertest.cc: Added test case for `term NOT "a phrase'. Thu May 06 00:28:32 BST 2004 Olly Betts <olly@survex.com> * common/omdebug.cc,common/omdebug.h: Renamed OmTimer class used internally by the --enable-debug=profile code to Xapian::Internal::Timer. Wed May 05 13:59:47 BST 2004 Olly Betts <olly@survex.com> * docs/scalability.html: Reworded to remove use of first person. Mon May 03 21:42:05 BST 2004 Olly Betts <olly@survex.com> * Makefile.am: Improved wording of a FIXME comment. Sun May 02 10:18:55 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Improved a couple of comments. Sat May 01 04:42:43 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h, backends/quartz/quartz_table.cc: Reopening the readonly version of a writable Btree is now more efficient. Sat May 01 01:41:24 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Added up-to-date success reports for x86-openbsd and x86-solaris. Fri Apr 30 04:05:15 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/: Don't delete and reload the Btree_base just after saving it - instead reuse the existing Btree_base object. Fri Apr 30 02:41:50 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Clean up new code in Btree::commit() from last change. Thu Apr 29 22:41:05 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/quartz_table.cc: Don't close and reopen the database table file descriptors every time we flush changes. Thu Apr 29 22:40:36 BST 2004 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Improved wording in a couple of places. Thu Apr 29 16:50:45 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/bcursor.cc: Removed references to Btree::error from debug code since Btree::error was removed a week ago. Tue Apr 27 20:46:25 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,configure.in: Added configure test for glibc, because otherwise we need to include a header before we can check for glibc in order to define something we should be defining before we include any headers! Tue Apr 27 17:45:51 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Only defined _XOPEN_SOURCE if __GLIBC__ is defined. OpenBSD seems to do the opposite to Linux and *disable* pread and pwrite if this is defined! Tue Apr 27 14:35:47 BST 2004 Olly Betts <olly@survex.com> * api/omenquire.cc,tests/api_db.cc: Need to adjust index by firstitem when indexing into items (bug#28). Mon Apr 26 23:45:15 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Successfully built with gcc-3.5-20040327 snapshot on x86 linux. Fri Apr 23 11:26:07 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Successfully built and tested on arm linux. Fri Apr 23 11:24:23 BST 2004 Olly Betts <olly@survex.com> * tests/remotetest.cc: Use 127.0.0.1 instead of localhost so that tcpmatch1 doesn't fail just because the network setup is broken. Thu Apr 22 00:54:58 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/: Stripped out Btree_errors as it's almost totally unused now. Fixed up the last few uses. * backends/quartz/quartz_table.cc: Fixed another case where read and write Btrees should share fds. Wed Apr 21 02:51:08 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Check the return value of fdatasync()/fsync()/_commit() and raise an error. They aren't likely to fail harmlessly. Wed Apr 21 02:45:28 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h, backends/quartz/quartz_table.cc: Share file descriptors between the read and write Btree objects so that a quartz WritableDatabase now uses 5 fds rather than 10. Wed Apr 21 02:37:38 BST 2004 Olly Betts <olly@survex.com> * docs/scalability.html: Removed the reference to the Quartz update bottleneck "currently being addressed for Xapian 0.8" as it's now been addressed! Wed Apr 21 00:24:08 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Corrected a comment. Mon Apr 19 14:42:06 BST 2004 Olly Betts <olly@survex.com> * NEWS: Final update for 0.8.0. Mon Apr 19 14:02:33 BST 2004 Olly Betts <olly@survex.com> * AUTHORS: Added people who've contributed build reports for the PLATFORMS file. Mon Apr 19 14:01:38 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: More updates from tinderbox and elsewhere. Sat Apr 17 02:24:11 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Lots of updates from tinderbox and mailing list. Sat Apr 17 00:13:26 BST 2004 Olly Betts <olly@survex.com> * HACKING: Added notes about using "using", and pointers to a couple of useful C++ web resources. Mon Apr 12 00:56:04 BST 2004 Olly Betts <olly@survex.com> * Makefile.am,NEWS,configure.in,extra/Makefile.am: Version 0.8.0. Mon Apr 12 00:31:15 BST 2004 Olly Betts <olly@survex.com> * NEWS: Updated for 0.8.0. Sun Apr 11 21:18:47 BST 2004 Olly Betts <olly@survex.com> * extra/queryparser.yy,extra/queryparsertest.cc: Don't use a raw term for a term which starts with a digit. Sat Apr 10 17:20:23 BST 2004 Olly Betts <olly@survex.com> * AUTHORS: "Open Muscat" not "OmSee"; Updated the list of contributors (more work needed...) Sat Apr 10 16:41:28 BST 2004 Olly Betts <olly@survex.com> * m4/xapian.m4: Make XO_LIB_XAPIAN with no arguments do what most users will want: check for Xapian and fail if it's not found, or AC_SUBST XAPIAN_CXXFLAGS and XAPIAN_LIBS if it is. Those will unusual needs can supply one or both arguments as at present. Sat Apr 10 15:00:28 BST 2004 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Sat Apr 10 00:32:07 BST 2004 Olly Betts <olly@survex.com> * PLATFORMS: Added success report for GCC 3.4 (prerelease). Sat Apr 10 00:27:58 BST 2004 Olly Betts <olly@survex.com> * extra/xapian/queryparser.h: Added doxygen documentation comments. Sat Apr 10 00:23:13 BST 2004 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Removed a now superfluous line. Sat Apr 10 00:21:15 BST 2004 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Give a more accurate message if valgrind spots a test doing something dodgy (e.g. free() of memory allocated with new). Thu Apr 08 17:22:20 BST 2004 Olly Betts <olly@survex.com> * net/.cvsignore: No longer need to ignore readquery.cc. Thu Apr 08 17:19:43 BST 2004 Olly Betts <olly@survex.com> * include/xapian/base.h: Note in doxygen comment why the ref_count is mutable. Thu Apr 08 15:33:16 BST 2004 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: valgrinding code needs errno.h too. Thu Apr 08 15:28:44 BST 2004 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Make Xapian::Weight::Weight() protected rather than private as we want to be able to call it from derived classes (GCC 3.4 flags this, other compilers seem to miss it). Tue Apr 06 03:32:39 BST 2004 Olly Betts <olly@survex.com> * net/omerr_string.cc,include/xapian/error.h, include/xapian/errortypes.h,include/xapian/output.h: Fixed some doxygen warnings. Mon Apr 05 17:21:06 BST 2004 Olly Betts <olly@survex.com> * README: There never was an Omsee release - the last BrightStation release was "OpenMuscat 0.4.1". Mon Apr 05 17:19:38 BST 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc, backends/quartz/quartz_database.h: Make the flush criterion 1000 documents changed (added, removed, or replaced) as that seems to perform much better over a wide variety of document sizes. Thu Apr 01 01:04:17 BST 2004 Olly Betts <olly@survex.com> * tests/quartztest.cc: Fixed temporary directory used (a slash was missing). Thu Apr 01 00:39:59 BST 2004 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Fri Mar 26 22:33:30 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Fix problems with termfreq and collfreq in postlist getting out of step when a recently modified or deleted document is deleted or remodified. Fri Mar 26 12:31:24 GMT 2004 Olly Betts <olly@survex.com> * configure.in: Fixed check for --enable-runtime-pseudo-reloc. Fri Mar 26 12:31:00 GMT 2004 Olly Betts <olly@survex.com> * bin/Makefile.am: Removed spurious "-lz" from quartzdump_LDADD. Thu Mar 25 00:22:13 GMT 2004 Olly Betts <olly@survex.com> * bin/quartzcheck.cc: Check the structure with the postlist Btree as well as the Btree structures themselves. Sun Mar 21 00:00:59 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_table_manager.cc: Fix for building on mingw. Sat Mar 20 23:04:27 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/positioniterator.h,include/xapian/postingiterator.h: Add doxygen comments in front of PostingIterator and PositionListIterator classes so that doxygen pulls out the existing documentation for their methods. Sat Mar 20 16:23:13 GMT 2004 Olly Betts <olly@survex.com> * configure.in,backends/quartz/quartz_table_manager.cc,common/utils.h: Added configure test for link() to avoid infinite loop on mingw! Thu Mar 18 12:20:18 GMT 2004 Olly Betts <olly@survex.com> * configure.in: Improved the --enable-runtime-pseudo-reloc check. Thu Mar 18 00:45:15 GMT 2004 Olly Betts <olly@survex.com> * bin/Makefile.am,tests/Makefile.am: Pass the magic mingw/cygwin flag when linking the programs in bin and tests. Thu Mar 18 00:33:22 GMT 2004 Olly Betts <olly@survex.com> * configure.in,xapian-config.in: mingw and cygwin both need -Wl,--enable-runtime-pseudo-reloc passing when linking. Arrange for xapian-config to include this, and check that the ld installed is a new enough version (or at least that it was at configure time). Tue Mar 16 01:40:43 GMT 2004 Olly Betts <olly@survex.com> * docs/matcherdesign.html: Merged in more details from a message I sent to the mailing list. Mon Mar 15 20:39:21 GMT 2004 Olly Betts <olly@survex.com> * tests/quartztest.cc: Fail the test if there's still a directory after rmdir(), or there isn't a directory after mkdir(). Close a QuartzDatabase or QuartzWritableDatabase before trying to overwrite it - cygwin doesn't allow use to delete open/locked files... Mon Mar 15 20:38:11 GMT 2004 Olly Betts <olly@survex.com> * backends/database.cc: Database::Internal::Internal::keep_alive() should be Database::Internal::keep_alive(). Mon Mar 15 17:20:58 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/base.h: Returning a void result makes SGI's compiler complain - don't! Mon Mar 15 14:34:44 GMT 2004 Olly Betts <olly@survex.com> * common/omdebug.cc: Only use O_SYNC (on the debug log) if we have it. Sun Mar 14 17:48:52 GMT 2004 Olly Betts <olly@survex.com> * tests/runtest.in: Mark with @configure_input@. Thu Mar 11 23:53:43 GMT 2004 Olly Betts <olly@survex.com> * common/database.h,common/termlist.h,include/xapian/database.h, include/xapian/enquire.h: Attempted fixes for warnings given by SGI's MIPSpro C++ compiler. Thu Mar 11 23:51:10 GMT 2004 Olly Betts <olly@survex.com> * configure.in: GCC 2.95 supported -Wno-long-long and is our minimum recommended version, so unconditionally use -Wno-long-long with GCC, and don't test for it on other compilers (the test incorrectly decided to use it with SGI's compiler leading to a warning for every file compiled). Thu Mar 11 17:29:24 GMT 2004 Olly Betts <olly@survex.com> * docs/intro_ir.html: Added a link to "Information Retrieval" by Keith v.R. which can be read on his website! Thu Mar 11 17:12:35 GMT 2004 Olly Betts <olly@survex.com> * PLATFORMS: Added IRIX success reports from Jim Lynch. Thu Mar 11 12:17:21 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Fixes for Sun C++'s fussy template matching. Thu Mar 11 12:12:46 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Another friend fix for Sun's C++. Thu Mar 11 01:56:19 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/query.h: Another fix for Sun's C++. Tue Mar 09 21:02:08 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/document.h: string -> std::string. Tue Mar 09 19:04:48 GMT 2004 Olly Betts <olly@survex.com> * api/omdocument.cc,common/document.h,include/xapian/document.h, tests/api_db.cc,tests/api_posdb.cc: Renamed Xapian::Document::add_term_nopos to Xapian::Document::add_term (with forwarding wrapper method for compatibility with existing code). Sat Mar 06 02:32:58 GMT 2004 Olly Betts <olly@survex.com> * docs/: Updated the quickstart tutorial and removed the warning that "this document isn't up to date". Sat Mar 06 01:56:17 GMT 2004 Olly Betts <olly@survex.com> * api/omenquire.cc: Another tweak for Sun's C++ compiler. Sat Mar 06 01:31:29 GMT 2004 Olly Betts <olly@survex.com> * common/omenquireinternal.h: Tweaked friend class declarations a bit so Sun's C++ compiler can cope. Thu Mar 04 23:58:17 GMT 2004 Olly Betts <olly@survex.com> * PLATFORMS: Updated with results from the tinderbox. Thu Mar 04 23:29:28 GMT 2004 Olly Betts <olly@survex.com> * tests/btreetest.cc,tests/quartztest.cc: NetBSD mkdir() doesn't cope with a trailing / on the path - fixed our code to cope with this. Wed Mar 03 19:14:09 GMT 2004 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Fri Feb 13 23:56:44 GMT 2004 Olly Betts <olly@survex.com> * api/omquery.cc,tests/api_nodb.cc: Throw error when an empty query is used to build in the binary operator Query ctor. Added regression test. Fri Feb 13 15:08:09 GMT 2004 Olly Betts <olly@survex.com> * HACKING: XAPIAN_DEBUG_TYPES should be XAPIAN_DEBUG_FLAGS. Sat Feb 07 14:26:54 GMT 2004 Olly Betts <olly@survex.com> * docs/queryparser.html: Grammar fixes. Fri Jan 16 02:06:53 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Fixed bug flagged up by deldoc4. Thu Jan 15 01:00:48 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_termlist.cc: Use Xapian::doccount instead of unsigned int in set_entries(). Thu Jan 15 00:56:47 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h, testsuite/btreecheck.cc: Made some static functions into static member functions of Btree so we can avoid duplicating code in btreecheck.cc. Also tweaked compare_keys to use memcmp and special case when the keys are the same length (each of these changes gives a very small speed gain). Thu Jan 15 00:54:47 GMT 2004 Olly Betts <olly@survex.com> * api/maptermlist.h: Removed a FIXME which already had been fixed. Removed a couple of asserts from a function which should never be called (and so had `Assert(false)' anyway). Mon Jan 12 00:32:37 GMT 2004 Olly Betts <olly@survex.com> * bin/quartzdump.cc: Backslash escape space and backslash in output rather than hex encoding them; renamed start-term and end-term to start-key and end-key; removed rather pointless "Calling next" message; if there's an error, write it to stderr not stdout, and exit with return code 1. Sun Jan 11 03:23:17 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_termlist.cc: When possible, pack the wdf into the same byte as the reuse length - doing so typically makes the termlist 14% smaller! This change is backward compatible. Sat Jan 10 03:04:22 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Fixed bug in new postlist chunking code. Sat Jan 10 00:45:04 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_termlist.cc: Removed unused OLD_TERMLIST_FORMAT code. Fri Jan 09 21:54:34 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Fix incorrect code which just happens to build and work with GCC 2.95. Fri Jan 09 14:40:00 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Tidied the initial call to get_chunk() out of the merge loop (gives ~4% speedup!) Fri Jan 09 01:31:05 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Implement chunking of postlists once again. Thu Jan 08 21:06:37 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Shortcut the very common case of appending to a posting list (which happens when we're just adding new documents). Declare small helper functions as "inline". Wed Jan 07 23:21:59 GMT 2004 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Set threshold for flushing to a saner value for the value we are now using. * backends/quartz/: Use freq_deltas to update the stats in the first chunk of each postlist. Mon Jan 05 16:00:07 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/enquire.h: Made the TradWeight constructor explicit. This is technically an API change as before you could pass a double where a Xapian::Weight was required - now you must pass Xapian::TradWeight(2.0) instead of 2.0. That seems desirable, and it's unlikely any existing code will be affected. Mon Jan 05 15:08:26 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/: Added "explicit" qualifier to internal ctors which take a single parameter. Mon Jan 05 15:04:12 GMT 2004 Olly Betts <olly@survex.com> * include/xapian/base.h: Assigning a normal pointer to a RefCntPtr no longer creates a temporary RefCntPtr from it. Thu Dec 25 05:49:03 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Changed to merge a batch of changes into a posting list in one pass. Thu Dec 25 05:48:09 GMT 2003 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Some minor improvements. Thu Dec 25 05:39:57 GMT 2003 Olly Betts <olly@survex.com> * tests/api_db.cc: Check returned docids are the expected values in a couple more cases. Improved wording of a comment. Thu Dec 25 05:38:43 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_utils.h: Updated explanation of string encoding. Thu Dec 25 05:37:36 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_values.cc: Fixed problem with dereferencing a pointer to the end of a string in debug output. Sun Dec 21 23:35:49 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Refactor recently relocated code to reduce duplication. Sun Dec 21 21:44:07 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Renamed skip_and_check_tname_in_key() to check_tname_in_key() as it doesn't do any sort of skipping! Sun Dec 21 14:54:49 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc,backends/quartz/quartz_postlist.h, tests/quartztest.cc: Subsumed QuartzPostList::add_entry and QuartzPostList::delete_entry into QuartzPostList::merge_changes. Sun Dec 21 03:21:42 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc, backends/quartz/quartz_postlist.cc, backends/quartz/quartz_postlist.h: Move merging on postlist changes into QuartzPostList in preparation for an efficient reimplementation. Sun Dec 21 01:47:07 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.h, backends/quartz/quartz_table_entries.h: Fixed typos and incorrect comments. Sat Dec 20 22:11:26 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/: Removed several needless inclusions of quartz_table_entries.h. Sat Dec 20 16:31:46 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/,docs/quartzdesign.html: Removed all the quartz lexicon code and docs. It's been disabled for ages, and we've not missed it. Sat Dec 20 01:21:47 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.h: Added doxygen comments for the QuartzWritableDatabase members we use to buffer changes. Sat Dec 20 01:19:10 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Trigger autoflush on totlen_added + totlen_removed rather than specially tracking the number of document add/delete/replace operations. Fri Dec 19 22:51:26 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/: Buffer up changes to the postlists and apply them all at once (at present they're applied inefficiently, but this is change enables us to apply them with an efficient merge). * tests/api_db.cc,tests/quartztest.cc: Added a couple of tests, and commented out some test lines which fail in debug builds. Sun Dec 14 03:38:55 GMT 2003 Olly Betts <olly@survex.com> * common/omdebug.cc: Open debug log with flag O_WRONLY so that we can actually write to it! Mon Dec 08 01:53:24 GMT 2003 Olly Betts <olly@survex.com> * HACKING: Update to mention that building from CVS requires ./configure --enable-maintainer-mode (or use bootstrap). Mon Dec 01 19:03:00 GMT 2003 Olly Betts <olly@survex.com> * configure.in: Fixed the "fixed" valgrind test. Sun Nov 30 23:42:52 GMT 2003 Olly Betts <olly@survex.com> * configure.in,testsuite/testsuite.cc: Fix test for valgrind - it wasn't working correctly when valgrind was installed but was too a version to support VALGRIND_COUNT_ERRORS and VALGRIND_COUNT_LEAKS. Wed Nov 26 16:31:19 GMT 2003 Olly Betts <olly@survex.com> * Makefile.am,extra/Makefile.am: Updated libtool -version-info parameters for 0.7.5. Wed Nov 26 15:32:45 GMT 2003 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.in: Version 0.7.5. Wed Nov 26 15:31:56 GMT 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy: Special case stemming language "none". Tue Nov 25 04:37:50 GMT 2003 Olly Betts <olly@survex.com> * api/ompositionlistiterator.cc,api/ompostlistiterator.cc, include/xapian/positioniterator.h,include/xapian/postingiterator.h, tests/api_db.cc: Added missing default ctors for PostingIterator and PositionIterator classes; fixed PositionIterator assignment operator. Tue Nov 25 03:40:11 GMT 2003 Olly Betts <olly@survex.com> * tests/btreetest.cc: Fixed 2 compiler warnings. Tue Nov 25 02:47:16 GMT 2003 Olly Betts <olly@survex.com> * common/,include/xapian/: Fixed incorrect doxygen comments which resulted in some missing text in the collated API and internal classes documentation. Fri Nov 21 03:23:57 GMT 2003 Olly Betts <olly@survex.com> * Makefile.am,include/xapian/termiterator.h,api/omtermlistiterator.cc: TermIterator::TermIterator() fix which actually works! Fri Nov 21 01:14:23 GMT 2003 Olly Betts <olly@survex.com> * HACKING,configure.in,docs/Makefile.am,extra/Makefile.am: Change to using AM_MAINTAINER_MODE. If you're doing development work on Xapian, you should configure with "--enable-maintainer-mode" and ideally use GNU make. Fri Nov 21 01:13:25 GMT 2003 Olly Betts <olly@survex.com> * include/xapian/termiterator.h,tests/api_nodb.cc: Added TermIterator default ctor which had disappeared somehow. Added regression test. Thu Nov 20 03:19:37 GMT 2003 Olly Betts <olly@survex.com> * configure.in: Fixed test for fdatasync to work (I suspect a change in a recent autoconf broke it as it relied on autoconf internal naming). Tue Nov 18 17:53:29 GMT 2003 Olly Betts <olly@survex.com> * extra/Makefile.am: Tweaked workaround for FreeBSD make in VPATH builds. Tue Nov 18 17:53:01 GMT 2003 Olly Betts <olly@survex.com> * extra/queryparsertest.cc: Added another testcase. Tue Nov 18 17:37:20 GMT 2003 Olly Betts <olly@survex.com> * docs/: Added first cut of documentation for Xapian::QueryParser query syntax. * docs/scalability.html: Fixed typo. Fri Oct 31 18:17:13 GMT 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy,extra/queryparsertest.cc: Added support for term prefixes on phrases and expressions (e.g. author:(twain OR poe) subject:"space flight"). Tue Oct 28 15:03:01 GMT 2003 Olly Betts <olly@survex.com> * api/omenquire.cc: Added FIXME comment. Tue Oct 28 15:01:14 GMT 2003 Olly Betts <olly@survex.com> * docs/todo.xml: Removed Java from list of language bindings we'd like to support, as we should have up-to-date Java bindings soon. Sun Oct 26 20:18:24 GMT 2003 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Fix the name that the meta file gets copied to (was /path/to/dbdirmeta rather than /path/to/dbdir/meta). Tue Oct 21 22:26:51 BST 2003 Olly Betts <olly@survex.com> * extra/Makefile.am: Backed out previous change - the original rule does work, but the build machine had a generated file which stopped the symlink being generated. Tue Oct 21 22:14:22 BST 2003 Olly Betts <olly@survex.com> * extra/Makefile.am: Tweaked rule to symlink queryparser.cc. Tue Oct 21 17:31:10 BST 2003 Olly Betts <olly@survex.com> * extra/Makefile.am: Added rule to symlink queryparser.cc from source tree to build tree - otherwise FreeBSD's make insists on regenerating it... Wed Oct 08 14:04:44 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/.cvsignore,tests/Makefile.am,testsuite/.cvsignore: Updated to reflect move of libbtreecheck.la from backends/quartz to testsuite. Wed Oct 08 13:22:08 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Don't pass mode argument to open when we aren't specifying O_CREAT. Thu Oct 02 15:29:38 BST 2003 Olly Betts <olly@survex.com> * NEWS: Another entry for omega for 0.7.4. Thu Oct 02 14:57:25 BST 2003 Olly Betts <olly@survex.com> * configure.in,extra/Makefile.am: Version 0.7.4. Thu Oct 02 14:56:36 BST 2003 Olly Betts <olly@survex.com> * HACKING: Improved wording in a few places. Thu Oct 02 14:39:49 BST 2003 Olly Betts <olly@survex.com> * NEWS: Updated. Thu Oct 02 13:35:46 BST 2003 Olly Betts <olly@survex.com> * xapian.spec.in: Don't hardcode the shared library versioning info. Wed Oct 01 13:51:15 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated - Solaris 8 + Sun Workshop C++ doesn't actually pass "make check" (I forgot the tinderbox builds omega before "make check"). Tue Sep 30 21:39:20 BST 2003 Olly Betts <olly@survex.com> * configure.in: Another stab at fixing the test for -lsocket. Tue Sep 30 18:29:31 BST 2003 Olly Betts <olly@survex.com> * configure.in: Quote arguments to test in a couple of places. Tue Sep 30 11:20:09 BST 2003 Olly Betts <olly@survex.com> * configure.in: Fixed test for -lsocket. Tue Sep 30 10:17:09 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated - Solaris 8 + Sun Workshop C++ builds and passes "make check" (but Omega fails to build). Mon Sep 29 00:41:19 BST 2003 Olly Betts <olly@survex.com> * configure.in: Fixed dependency libs for libxapian. Sun Sep 28 23:55:49 BST 2003 Olly Betts <olly@survex.com> * Makefile.am,backends/quartz/Makefile.am, backends/quartz/btreecheck.cc,backends/quartz/btreecheck.h, bin/Makefile.am,testsuite/Makefile.am,testsuite/btreecheck.cc, testsuite/btreecheck.h: libbtreecheck.la now has an explicit dependency on libxapian.la. Sun Sep 28 17:29:55 BST 2003 Olly Betts <olly@survex.com> * Makefile.am,configure.in: Tell libtool about any libraries which libxapian needs to depend on. Fri Sep 26 09:39:29 BST 2003 Olly Betts <olly@survex.com> * bin/Makefile.am: Swap order of libbtreecheck.la and libxapian.la when linking quartzcheck to see if that gets it to work with Sun's CC. Thu Sep 25 17:25:17 BST 2003 Olly Betts <olly@survex.com> * xapian-config.in: Fixed stupid mistake in previous check-in. Thu Sep 25 16:50:00 BST 2003 Olly Betts <olly@survex.com> * xapian-config.in: Link with libxapianqueryparser before libxapian, since that's the dependency order. Thu Sep 25 16:43:20 BST 2003 Olly Betts <olly@survex.com> * include/om/om.h: Updated comment about why we `#include <iostream>'. Thu Sep 25 13:13:37 BST 2003 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,backends/database.cc,common/omdebug.cc, common/omstringstream.h,include/xapian/output.h: Removed or replaced uses of <iostream> and <iosfwd> - we don't need or want the library to pull in cin and friends. Thu Sep 25 13:12:31 BST 2003 Olly Betts <olly@survex.com> * Makefile.am: Improved several comments. Thu Sep 25 11:57:03 BST 2003 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Removed `#include <iostream>' left over from debugging; changed to `using namespace std;'. Mon Sep 15 00:41:10 BST 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy: Fixed to build with Sun's C++ compiler. Thu Sep 11 03:35:29 BST 2003 Olly Betts <olly@survex.com> * dummy.c,dummy.cc,Makefile.am: Change the default file to be C++ so that automake tells libtool that this is a C++ library. * Makefile.am: Pass -no-undefined to libtool so that it will build a DLL on MS Windows. Wed Sep 10 02:53:46 BST 2003 Olly Betts <olly@survex.com> * docs/scalability.html: Fixed oops (32GB should be 32TB); Added note about Linux 2.4 and ext2 filesize limits. Tue Sep 09 03:36:05 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: don't start the document's TermIterator from scratch on every iteration in replace_document(). Should be a small performance win. Sun Sep 07 22:17:27 BST 2003 Olly Betts <olly@survex.com> * configure.in: Fixed check for socketpair - we were automatically disabling the remote backend on platforms where socketpair is in libsocket (such as Solaris). Sun Sep 07 19:56:57 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Added success report for cygwin from Eric B. Ridge. Sun Sep 07 03:53:48 BST 2003 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Also need to rerun test if it "probably leaked". Sun Sep 07 03:51:37 BST 2003 Olly Betts <olly@survex.com> * tests/api_db.cc: Use x & (x - 1) trick to check for x being a power of 2, rather than an elaborate loop with shifting; use Xapian::docid rather than "int" and "unsigned int". Sun Sep 07 02:24:06 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc: Fixed small memory leak if Xapian::Enquire::set_query() is called more than once. Sun Sep 07 02:12:47 BST 2003 Olly Betts <olly@survex.com> * common/omenquireinternal.h,include/xapian/enquire.h: Fixed Xapian::ESet to have reference counted internals * Makefile.am: bumped libxapian version-info to 1:0:0 because of this. Sun Sep 07 02:09:50 BST 2003 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Delete any buffered test output before asking valgrind to check for leaks, as otherwise the buffered output may be reported as a leak; also fixed |= to != in several places - a cosmetic bug which lead to extra newlines after test output. Sun Sep 07 00:09:32 BST 2003 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Actually open temporary file for valgrind to log to. Sat Sep 06 20:50:58 BST 2003 Olly Betts <olly@survex.com> * configure.in,testsuite/testsuite.cc: Fixed up detection of valgrind and testsuite code to use it. Sat Sep 06 20:48:17 BST 2003 Olly Betts <olly@survex.com> * tests/btreetest.cc: Reinstated non-superfluous `/'s which I had overeagerly removed! Sat Sep 06 17:28:00 BST 2003 Olly Betts <olly@survex.com> * matcher/Makefile.am: Ship networkmatch.cc even if "make dist" is run from a tree with the remote backend disabled. Thu Aug 28 00:01:22 BST 2003 Olly Betts <olly@survex.com> * tests/btreetest.cc: Removed superfluous `/'s from constructed paths. Fri Aug 15 15:12:24 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated with current testsuite results for mingw build. Fri Aug 15 15:10:59 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Use O_BINARY for database I/O if it exists. Fri Aug 15 15:10:30 BST 2003 Olly Betts <olly@survex.com> * common/utils.h: mkdir() only takes one argument on mingw. Fri Aug 15 01:26:22 BST 2003 Olly Betts <olly@survex.com> * common/utils.h,testsuite/backendmanager.cc: Touch file using open() rather than system(). Fri Aug 15 01:06:06 BST 2003 Olly Betts <olly@survex.com> * common/utils.h: Fixed parameter name of system from "filename" to "command" (cosmetic change). Thu Aug 14 22:31:23 BST 2003 Olly Betts <olly@survex.com> * net/progclient.cc: Tidied a little. Tue Aug 12 12:50:19 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Pass 0 for the lexicon/postlist table when creating a termlist just to find the doc length. Tue Aug 12 02:08:50 BST 2003 Olly Betts <olly@survex.com> * api/maptermlist.h,api/omdocument.cc,common/documentterm.h: Removed unused OmDocumentTerm::termfreq member variable. Tue Aug 12 00:53:02 BST 2003 Olly Betts <olly@survex.com> * api/omdocument.cc,common/documentterm.h: OmDocumentTerm ctor now takes wdf, and replaced set_wdf() with inc_wdf() and dec_wdf(). Mon Aug 11 16:11:26 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_table_entries.cc: Removed rather unnecessary use of const_cast. Mon Aug 11 16:11:05 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_table.cc: Removed unused variable. Mon Aug 11 16:09:57 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Fixed typo in comment. Mon Aug 11 15:51:20 BST 2003 Olly Betts <olly@survex.com> * tests/quartztest.cc: adddoc2 now checks that there weren't any extra values created. Mon Aug 11 12:55:49 BST 2003 Olly Betts <olly@survex.com> * configure.in: Sorted out tests for gethostbyname and gethostbyaddr using standard autoconf macros. Mon Aug 11 12:05:13 BST 2003 Olly Betts <olly@survex.com> * common/match.h,matcher/: Removed unused open_document() method from SubMatch and derived classes; calls made by the matcher to Document::Internal::open_document() now use the lazy flag provided for precisely this purpose, but apparently never used - this should give quite a speed boost to any matcher options which use values (e.g. sort, collapse). Sun Aug 10 02:25:59 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Out of curiosity, I tried to build with a Linux->DJGPP cross-compiler I had to hand. It built (just had to disable the remote backend if socketpair isn't found and add the workaround for not having snprintf). Sun Aug 10 01:38:17 BST 2003 Olly Betts <olly@survex.com> * common/utils.cc: Fixed to compile if snprintf isn't available. Sun Aug 10 01:37:44 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Fixed "initialisation reordered" warning. Sun Aug 10 01:26:58 BST 2003 Olly Betts <olly@survex.com> * acinclude.m4: Updated in line with previous change. Sun Aug 10 01:00:08 BST 2003 Olly Betts <olly@survex.com> * configure.in: If fork is found, but socketpair isn't, automatically disable the remote backend rather than configure dying with an error. * autoconf/: Removed various unused autoconf macros. Sat Aug 09 23:57:44 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btree.h: Made "overwritten" protected. Sat Aug 09 23:45:26 BST 2003 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Wrapped an overlong line. Sat Aug 09 23:31:30 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h, backends/quartz/quartz_table.cc,docs/quartzdesign.html: Improved class Btree's encapsulation. Sat Aug 09 22:30:13 BST 2003 Olly Betts <olly@survex.com> * NEWS: Fixed a few typos. Sat Aug 09 22:29:52 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated with results from the Tinderbox. Sat Aug 09 22:19:12 BST 2003 Fabrice Colin * xapian.spec.in: Updated RPM packaging to reflect changes in 0.7.3. Fri Aug 08 01:38:40 BST 2003 Olly Betts <olly@survex.com> * Makefile.am,configure.in,extra/Makefile.am: Version 0.7.3. Fri Aug 08 01:35:04 BST 2003 Olly Betts <olly@survex.com> * NEWS: Updated. Fri Aug 08 01:34:29 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated with more tinderbox results. Wed Aug 06 16:27:34 BST 2003 Olly Betts <olly@survex.com> * HACKING: Noted that libtool 1.5 is the first version to actually support linking C++ libraries properly. Wed Aug 06 16:25:15 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc,net/socketcommon.cc,common/omenquireinternal.h: Moved rset serialisation into a method of RSet::Internal, so omrset_to_string() is now just glue code. This eliminates the need for it to be a friend of RSet::Internal which Sun's C++ compiler didn't seem to be able to cope with. Wed Aug 06 01:57:11 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc: Applied patch from Olivier Galibert to fix MSetIterator::get_document() when get_mset() was called with first != 0. * tests/api_db.cc: Added regression test (msetiterator3). Mon Aug 04 12:40:25 BST 2003 Olly Betts <olly@survex.com> * languages/Makefile.am: Fixed compilation for compilers other than GCC. Wed Jul 30 17:03:14 BST 2003 Olly Betts <olly@survex.com> * common/getopt.h: Another try - the previous one failed on FreeBSD. Wed Jul 30 15:07:58 BST 2003 Olly Betts <olly@survex.com> * common/getopt.h: Another iteration of the getopt C++ prototype fix as the previous one failed on Solaris. Wed Jul 30 02:38:47 BST 2003 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc: Fixed breakage caused by remotetest cleanup. Tue Jul 29 13:22:20 BST 2003 Olly Betts <olly@survex.com> * tests/remotetest.cc: Cleaned up uses of BackendManager::get_database() to pass a string rather than a vector with one string in. Tue Jul 29 11:58:21 BST 2003 Olly Betts <olly@survex.com> * testsuite/testutils.cc: Added output of const vector<unsigned int> for platforms which use our internal stringstream implementation. Tue Jul 29 11:42:54 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated with results from the tinderbox. Tue Jul 29 01:21:53 BST 2003 Olly Betts <olly@survex.com> * testsuite/testsuite.cc,testsuite/testsuite.h: Only use \r in test output if the output is a tty. Mon Jul 28 23:56:10 BST 2003 Olly Betts <olly@survex.com> * xapian-config.in: Fixed fix of --cxxflags for uninstalled VPATH builds. Mon Jul 28 23:37:15 BST 2003 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc: Increased default timeout used by tests running on the remote backend from 10 seconds to 5 minutes to avoid tests failing just because the machine running them is slow and/or busy. Mon Jul 28 15:41:12 BST 2003 Olly Betts <olly@survex.com> * testsuite/testutils.h: Fixed check for broken exception handling - we were getting "Xapian::" prefixed to one version and not on the other. Mon Jul 28 15:34:54 BST 2003 Olly Betts <olly@survex.com> * tests/runtest.in: Set srcdir if it isn't already to make it easy to manually run test programs from a VPATH build. Mon Jul 28 10:38:04 BST 2003 Olly Betts <olly@survex.com> * xapian-config.in: Fixed --cxxflags for uninstalled VPATH builds. Mon Jul 28 10:37:17 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated FreeBSD success report. Sun Jul 27 16:01:25 BST 2003 Olly Betts <olly@survex.com> * configure.in: Explicitly remove include/xapian/version.h from APIDOC_SRC to avoid us trying to run doxygen unnecessarily. Sun Jul 27 14:31:36 BST 2003 Olly Betts <olly@survex.com> * docs/todo.xml: Assorted updates. Sun Jul 27 13:02:36 BST 2003 Olly Betts <olly@survex.com> * extra/.cvsignore: We still create libomqueryparser.la for backward compatibility. Sun Jul 27 04:56:40 BST 2003 Olly Betts <olly@survex.com> * xapian-config.in,m4/xapian.m4: Added --from-xo-lib-xapian option to xapian-config for use by XO_LIB_XAPIAN to allow us to give a more helpful error message if asked to link an uninstalled libxapian. Sun Jul 27 04:46:48 BST 2003 Olly Betts <olly@survex.com> * xapian-config.in: Fixed --cxxflags in non-installed case. Tidied up --help output and error messages. Sun Jul 27 01:40:39 BST 2003 Olly Betts <olly@survex.com> * tests/api_db.cc: Check termfreq in allterms4. Sun Jul 27 01:39:57 BST 2003 Olly Betts <olly@survex.com> * extra/queryparsertest.cc: Fixed for relocated queryparser.h. Sun Jul 27 00:59:56 BST 2003 Olly Betts <olly@survex.com> * .cvsignore,backends/quartz/.cvsignore,extra/.cvsignore: Updated. Sun Jul 27 00:19:11 BST 2003 Olly Betts <olly@survex.com> * tests/Makefile.am: Oops, failed to remove btreecheck.cc reference in previous check-in. Sat Jul 26 23:36:01 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/Makefile.am,bin/Makefile.am,tests/Makefile.am: Hopefully fixed "make dist" in VPATH builds. Sat Jul 26 22:58:27 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/Makefile.am: Need to distribute btreecheck.cc. Sat Jul 26 20:05:47 BST 2003 Olly Betts <olly@survex.com> * xapian-config.in,extra/Makefile.am,extra/queryparser.h, extra/queryparser.yy,extra/xapian/queryparser.h: Sorted out xapian-config so it handles using QueryParser too. Sat Jul 26 19:03:04 BST 2003 Olly Betts <olly@survex.com> * xapian-config.in: Updated logic in --libs and --cxxflags in line with recent change in --ltlibs. Sat Jul 26 18:31:11 BST 2003 Olly Betts <olly@survex.com> * xapian-config.in: Reworked logic for finding libxapian.la in --ltlibs - it may not yet exist in the non-installed case. * m4/xapian.m4: Don't pass --libs when checking if xapian-config works as it will fail in the non-installed case. Sat Jul 26 01:25:46 BST 2003 Olly Betts <olly@survex.com> * NEWS: Updated. Fri Jul 25 10:20:18 BST 2003 Olly Betts <olly@survex.com> * NEWS: Updated in preparation for 0.7.3 release. Thu Jul 24 23:20:14 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_alltermslist.cc, backends/quartz/quartz_alltermslist.h,tests/api_db.cc: Fixed allterms TermIterator to not give duplicate terms with a quartz database when a posting list is chunked; added regression test (allterms4). Mon Jul 21 15:06:59 BST 2003 Richard Boulton <richard@tartarus.org> * include/xapian/enquire.h: Fix incorrect documentation comment for Enquire::set_set_forward(). (Looked like a cut&paste error) Mon Jul 21 01:10:55 BST 2003 Olly Betts <olly@survex.com> * docs/install.html,m4/xapian.m4: Added XO_LIB_XAPIAN to replace OM_PATH_XAPIAN. XO_LIB_XAPIAN will automagically enable use of "xapian-config --ltlibs" if A[CM]_PROG_LIBTOOL is used in configure.in. Sun Jul 20 16:00:43 BST 2003 Olly Betts <olly@survex.com> * .cvsignore,Makefile.am,configure.in,xapian-config.in, xapian-config.nodep.in: Reworked xapian-config so that it now supports linking with libtool - using libtool means that the run-time library path is set and that you can now link with an uninstalled libxapian. Also xapian-config will now work once configure is run, rather than only after "make all". Thu Jul 17 01:42:47 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/bcursor.cc,backends/quartz/btree.cc: Check for EINTR when reading or writing blocks and retry the operation. Thu Jul 17 00:56:35 BST 2003 Olly Betts <olly@survex.com> * COPYING: Updated FSF address, and reinstated missing section: "How to Apply These Terms to Your New Programs" Thu Jul 17 00:48:40 BST 2003 Olly Betts <olly@survex.com> * HACKING,bootstrap: Removed bootstrap in favour of top-level bootstrap. Mon Jul 14 23:36:14 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated some linux results: RH7.3 on x86, and Debian on alpha and arm. Mon Jul 14 23:33:17 BST 2003 Olly Betts <olly@survex.com> * configure.in,Makefile.am,extra/Makefile.am: Added versioning information to libxapian and libomqueryparser. Mon Jul 14 23:32:15 BST 2003 Olly Betts <olly@survex.com> * bin/quartzdump.cc,testsuite/testsuite.cc: Added long option support to quartzdump, and to the testsuite programs. Mon Jul 14 22:45:53 BST 2003 Olly Betts <olly@survex.com> * tests/internaltest.cc: Changed test exception1 to actually test something (hopefully what was originally intended!) Mon Jul 14 17:15:28 BST 2003 Olly Betts <olly@survex.com> * common/getopt.h: #include <stdlib.h> before defining getopt as a macro - hopefully that'll avoid problems with clobbering prototypes of getopt() in system headers. Sat Jul 12 09:15:26 BST 2003 Olly Betts <olly@survex.com> * bin/quartzcompact.cc: Need stdio.h for rename(). Sat Jul 12 01:11:54 BST 2003 Olly Betts <olly@survex.com> * HACKING: Noted another reason why libtool 1.5 is needed. Sat Jul 12 01:11:08 BST 2003 Olly Betts <olly@survex.com> * docs/mkdoc.pl: Don't choke on a comment at the end of the DIST_SUBDIRS line in a Makefile.am. Fri Jul 11 22:00:36 BST 2003 Olly Betts <olly@survex.com> * HACKING: Added note about additional tools needed for building a distribution. Fri Jul 11 15:13:08 BST 2003 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.in: Version 0.7.2. Fri Jul 11 12:36:18 BST 2003 Olly Betts <olly@survex.com> * NEWS: Updated in preparation for 0.7.2. Fri Jul 11 12:12:40 BST 2003 Olly Betts <olly@survex.com> * configure.in: Fixed test for snprintf. Fri Jul 11 09:37:14 BST 2003 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Fixed dereference of a NULL pointer which was happening when a test threw an unexpected exception. Fri Jul 11 09:31:40 BST 2003 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc: Backed out incorrect hunk of previous patch which was causing test failures. Fri Jul 11 01:48:21 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_table_manager.cc,tests/quartztest.cc, testsuite/backendmanager.cc: Tidying up around directory creation. Fri Jul 11 00:57:12 BST 2003 Olly Betts <olly@survex.com> * include/xapian/database.h: Document that the directory is created. Fri Jul 11 00:53:59 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_table_manager.cc: When asked to create a quartz database, try to create the directory if it doesn't already exist. Then we don't have to do it in every single Xapian program which wants to create a database... Fri Jul 11 00:36:00 BST 2003 Olly Betts <olly@survex.com> * common/getopt.h: Removed space between \ and newline. Fri Jul 11 00:32:40 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Added MacOS X 10.2 success report. Thu Jul 10 20:05:58 BST 2003 Olly Betts <olly@survex.com> * common/getopt.h: Fixed to work better with C++ compilers on non-glibc platforms. Thu Jul 10 19:21:00 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Modified setting of _XOPEN_SOURCE to actually work! Thu Jul 10 19:19:30 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Defined _XOPEN_SOURCE=500 for GLIBC so we get pread() and pwrite(). Thu Jul 10 00:28:25 BST 2003 Olly Betts <olly@survex.com> * HACKING,Makefile.am,configure.in,bindings/.cvsignore, bindings/Makefile.am,bindings/dir_contents,bindings/guile/.cvsignore, bindings/guile/Makefile.am,bindings/php4/.cvsignore, bindings/php4/ABOUT,bindings/php4/Makefile.am,bindings/python/, bindings/swig/,bindings/tcl8/.cvsignore,bindings/tcl8/Makefile.am, docs/todo.xml: Removed bindings - they'll shortly reappear in a separate module named xapian-bindings. Thu Jul 10 00:03:04 BST 2003 Olly Betts <olly@survex.com> * bindings/swig/om_util_perl5.i: Removed. Wed Jul 09 15:35:56 BST 2003 Olly Betts <olly@survex.com> * bindings/guile/Makefile.am,bindings/php4/Makefile.am, bindings/python/Makefile.am,bindings/swig/, bindings/tcl8/Makefile.am: Removed omtypes.i and merged omstem.i into xapian.i; added forward declaration for upwrapped classes to xapian.i. Wed Jul 09 15:34:11 BST 2003 Olly Betts <olly@survex.com> * bindings/java/,bindings/java/com/muscat/om/,bindings/java/native/: Removed old Java bindings - we'll do them using SWIG instead. Wed Jul 09 14:51:51 BST 2003 Olly Betts <olly@survex.com> * bindings/php4/Makefile.am,bindings/php4/xapian/.cvsignore: More work on the php4 bindings. Wed Jul 09 03:00:17 BST 2003 Olly Betts <olly@survex.com> * bindings/guile/Makefile.am,bindings/python/Makefile.am, bindings/tcl8/Makefile.am: Worked towards getting guile and tcl8 bindings up-to-date. * bindings/php4/.cvsignore: Ignore the whole xapian subdirectory. Wed Jul 09 02:37:12 BST 2003 Olly Betts <olly@survex.com> * bindings/perl5/: Removed remaining files. Wed Jul 09 02:30:55 BST 2003 Olly Betts <olly@survex.com> * configure.in,bindings/Makefile.am,bindings/php4/Makefile.am: Working towards getting the php4 bindings functional again. Wed Jul 09 01:27:39 BST 2003 Olly Betts <olly@survex.com> * bindings/swig/om_util_php4.i: om_termname -> std::string. Wed Jul 09 01:25:11 BST 2003 Olly Betts <olly@survex.com> * docs/overview.html,include/xapian/base.h,include/xapian/database.h: Improvements to doxygen-generated documentation. Wed Jul 09 01:23:55 BST 2003 Olly Betts <olly@survex.com> * configure.in,docs/todo.xml: Added check for SWIG version (require at least 1.3.14). Tue Jul 8 20:59:08 2003 James Aylett <james@tartarus.org> * common/utils.h: missing #include <ctype.h> Tue Jul 8 18:06:23 2003 James Aylett <james@tartarus.org> * bindings/swig/xapian.i: over-enthusiastic automatic conversion of termname to std::string Tue Jul 08 17:45:16 BST 2003 Olly Betts <olly@survex.com> * common/utils.h: Improved mingw implementation of rmdir(). Tue Jul 08 17:43:21 BST 2003 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.in: Version 0.7.1. Sun Jul 06 21:35:08 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Minor tweak. Sun Jul 06 21:34:30 BST 2003 Olly Betts <olly@survex.com> * common/utils.h: Compile fixes for mingw. Sun Jul 06 17:59:26 BST 2003 Olly Betts <olly@survex.com> * docs/scalability.html: Fixed typo. Sun Jul 06 13:02:09 BST 2003 Olly Betts <olly@survex.com> * docs/scalability.html: Fixed typo. Sun Jul 06 12:56:08 BST 2003 Olly Betts <olly@survex.com> * docs/Makefile.am,docs/index.html,docs/scalability.html: Added new document on Xapian's scalability. Sun Jul 06 12:15:34 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/: Made quartz block count unsigned, which should nearly double the size of database for a given block size. Sun Jul 06 11:47:46 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/: Use BLK_UNUSED rather than literal -1. Sun Jul 06 11:47:14 BST 2003 Olly Betts <olly@survex.com> * common/utils.h: Fixed last check-in to compile. Sun Jul 06 11:01:03 BST 2003 Olly Betts <olly@survex.com> * common/utils.h,tests/btreetest.cc,tests/quartztest.cc, testsuite/backendmanager.cc: Replaced system("rm -rf ...") with rmdir() function and added two ms windows implementations for Sam to try. Sat Jul 05 10:37:35 BST 2003 Olly Betts <olly@survex.com> * getopt/getopt.c: Suppress "unused variable" warnings from getopt.c. Sat Jul 05 02:21:11 BST 2003 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Sat Jul 05 02:09:06 BST 2003 Olly Betts <olly@survex.com> * configure.in,backends/quartz/btree.cc: Use pread() and pwrite() if available - this avoids one syscall per block read/write. Fri Jul 04 22:05:14 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Added results from building 0.7.0 on my Linux dev box. Fri Jul 04 18:52:07 BST 2003 Olly Betts <olly@survex.com> * extra/Makefile.am: Distribute symboltab.h. Fri Jul 04 18:04:53 BST 2003 Olly Betts <olly@survex.com> * common/Makefile.am: Distribute documentterm.h. Fri Jul 04 15:30:28 BST 2003 Olly Betts <olly@survex.com> * configure.in: Disabled guile, tcl8, and php4 here as well as in bindings/Makefile.am. Fri Jul 4 11:27:53 2003 James Aylett <james@tartarus.org> * bindings/Makefile.am: change commenting so mkdoc.pl doesn't choke Fri Jul 4 10:51:04 2003 James Aylett <james@tartarus.org> * bootstrap: add missing ';;' as case pattern delimiter Thu Jul 03 23:46:14 BST 2003 Olly Betts <olly@survex.com> * bindings/Makefile.am,bindings/python/Makefile.am: Fixed to work with "make dist". Thu Jul 03 23:29:06 BST 2003 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.in: Version 0.7.0. Thu Jul 03 22:57:22 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated (Linux GCC 2.95 and Linux to mingw cross). Thu Jul 03 22:55:54 BST 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy,extra/queryparsertest.cc: Allow more than one adjacent phrase generating character. Thu Jul 03 22:12:18 BST 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy,extra/queryparsertest.cc: Always ignore phrase generating characters at start and end of query. Thu Jul 03 21:02:10 BST 2003 Olly Betts <olly@survex.com> * extra/Makefile.am: Fixed for mingw compile. Thu Jul 03 19:24:50 BST 2003 Olly Betts <olly@survex.com> * buildall,docs/OMAims: Removed unwanted old files. Thu Jul 03 19:18:41 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/runquartztest: Obsolete, so removed. Thu Jul 03 01:26:03 BST 2003 Olly Betts <olly@survex.com> * common/omenquireinternal.h: Fix for building with GCC 3.3. Thu Jul 03 01:02:54 BST 2003 Olly Betts <olly@survex.com> * NEWS: Updated. Wed Jul 02 22:18:07 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/bcursor.cc,backends/quartz/bcursor.h: Made Bcursor::get_key(string * key) const. Wed Jul 02 21:51:43 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/: Workaround for shared_level problem turns out to be arguably the better approach, so made it permanent and tidied up code. Mon Jun 23 21:39:50 BST 2003 Olly Betts <olly@survex.com> * NEWS: Updated in preparation for 0.7.0 release. Mon Jun 23 21:18:23 BST 2003 Olly Betts <olly@survex.com> * extra/omparsequery.h: Corrected include protection token comment on #endif to match that used in the #ifdef. Mon Jun 23 21:10:33 BST 2003 Olly Betts <olly@survex.com> * bootstrap: Check for Bison 1.875 which doesn't work with Xapian. Mon Jun 23 21:08:35 BST 2003 Olly Betts <olly@survex.com> * HACKING: Automake 1.7 works reliably with Xapian; CFLAGS=-Werror can now be used as we suppress the warnings from the Snowball generated C code. Mon Jun 23 16:55:54 BST 2003 Olly Betts <olly@survex.com> * api/,backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h,backends/quartz/, bindings/swig/om_util_python.i,bindings/swig/xapian.i,common/, include/Makefile.am,include/xapian.h,include/om/om.h,include/xapian/, matcher/localmatch.h,matcher/phrasepostlist.h,tests/api_db.cc, tests/api_posdb.cc,tests/quartztest.cc: Xapian::Xapian::PostListIterator -> Xapian::PostingIterator; Xapian::PositionListIterator -> Xapian::PositionIterator; xapian/postlistiterator.h -> xapian/postingiterator.h; xapian/positionlistiterator.h -> xapian/positioniterator.h. Mon Jun 23 15:31:30 BST 2003 Olly Betts <olly@survex.com> * backends/database.cc,backends/inmemory/inmemory_database.h, backends/muscat36/,backends/quartz/,bindings/python/Makefile.am, bindings/swig/om_util.i,bindings/swig/omstem.i,docs/quickstart.html, docs/todo.xml,include/xapian/database.h,include/xapian/enquire.h, include/xapian/errortypes.h,tests/api_db.cc,tests/quartztest.cc, testsuite/backendmanager.cc,testsuite/textfile_indexer.cc: Xapian::OpeningError renamed to Xapian::DatabaseOpeningError. * include/om/om.h: OmOpeningError now maps to Xapian::DatabaseOpeningError. Mon Jun 23 01:44:45 BST 2003 Olly Betts <olly@survex.com> * api/,backends/inmemory/inmemory_database.cc, backends/muscat36/da_database.cc,backends/muscat36/db_database.cc, backends/net/net_database.cc,backends/net/net_termlist.cc, backends/net/net_termlist.h,backends/quartz/,bin/omprogsrv.cc, bin/quartzdump.cc,common/,include/xapian/error.h, include/xapian/errorhandler.h,matcher/emptymatch.h,net/, tests/quartztest.cc,tests/stemtest.cc,testsuite/testsuite.cc, testsuite/textfile_indexer.cc: Use '#include <...>' for headers in include/. Thu Jun 19 17:55:07 BST 2003 Olly Betts <olly@survex.com> * common/netserver.h,common/socketserver.h,matcher/localmatch.cc, matcher/multimatch.cc,net/socketserver.cc: Removed unused header inclusion; pass Stats and Xapian::Database by const reference, not value. Thu Jun 19 14:50:49 BST 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy,extra/symboltab.h: Force control characters to spaces; Added note that accent normalisation is perhaps done at the wrong time... Thu Jun 19 14:40:36 BST 2003 Olly Betts <olly@survex.com> * extra/queryparsertest.cc: Added test for control characters. Thu Jun 19 14:31:12 BST 2003 Olly Betts <olly@survex.com> * extra/queryparsertest.cc: Added checks for weird whitespace in queries. Wed Jun 18 19:02:14 BST 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy: Reworked parsing of boolean operators to improve error reporting. * extra/queryparsertest.cc: Added 2 more testcases and fixed to better handle tests for expected errors. Wed Jun 18 19:00:55 BST 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy: Trim leading and trailing whitespace before attempting to parse the query. Wed Jun 18 17:58:56 BST 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy: `using namespace Xapian;' Fri Jun 06 01:33:33 BST 2003 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Commented out the section on the Lexicon, which we no longer use. Thu Jun 05 22:40:21 BST 2003 Olly Betts <olly@survex.com> * bindings/python/.cvsignore,bindings/python/Makefile.am, bindings/python/dothemake.txt,bindings/swig/om_util_python.i: Cleaned up building of python bindings by making better use of automake. Thu Jun 05 22:17:40 BST 2003 Olly Betts <olly@survex.com> * languages/Makefile.am: Reenable standard warnings apart from the -Wunused. Thu Jun 05 20:21:08 BST 2003 Olly Betts <olly@survex.com> * api/omqueryinternal.cc: Fixed a couple of problems with the new query serialisation. Thu Jun 05 17:55:21 BST 2003 Olly Betts <olly@survex.com> * HACKING,configure.in: Flex is no longer used, so removed configure tests and note in HACKING. Thu Jun 05 17:53:30 BST 2003 Olly Betts <olly@survex.com> * tests/remotetest.cc: In test netexpand1, use TEST() rather than Assert() so that the check is made in non-debug builds too. Thu Jun 05 17:52:17 BST 2003 Olly Betts <olly@survex.com> * api/omqueryinternal.cc,common/socketcommon.h,docs/todo.xml, include/xapian/query.h,net/: Reworked query serialisation so that the code is now all in api/omqueryinternal.cc. Serialisation is now rather more compact and no longer relies on flex for parsing. Thu Jun 05 17:45:00 BST 2003 Olly Betts <olly@survex.com> * bin/Makefile.am: omprogsrv depends on libtest.la, so it need to be in check_PROGRAMS. Thu Jun 05 17:09:10 BST 2003 Olly Betts <olly@survex.com> * Makefile.am: Moved bindings after extra in SUBDIRS - the bindings wrap Xapian::QueryParser so they need to be built after it. Thu Jun 05 01:28:01 BST 2003 Olly Betts <olly@survex.com> * bindings/python/Makefile.am: Use $(SWIG) (as detected by configure). Thu Jun 05 00:46:16 BST 2003 Olly Betts <olly@survex.com> * configure.in,bindings/Makefile.am,bindings/python/Makefile.am: Check for python and find the correct paths for building python bindings; Added dependencies for building _xapian.so. Wed Jun 4 17:19:37 2003 James Aylett <james@tartarus.org> * configure.in, Makefile.am, bindings: build system will build bindings as part of normal build process. No longer build Perl or Java bindings; for Perl, use Search::Xapian; Java will be replaced with SWIG-based bindings. * bindings: bindings now use Xapian:: namespace names rather than the old Om* ones. Almost all of the API should now be wrapped. Wed Jun 04 02:52:38 BST 2003 Olly Betts <olly@survex.com> * docs/overview.html: Added start of "Design Principles" section. * docs/todo.xml: Added a couple more items. Wed Jun 04 02:51:56 BST 2003 Olly Betts <olly@survex.com> * Makefile.am: Build testsuite later - it now doesn't need to be built until just before tests. Wed Jun 04 02:43:08 BST 2003 Olly Betts <olly@survex.com> * backends/muscat36/: Split up spec into comments in the appropriate header files; fixed to build after recent Om -> Xapian changes. Wed Jun 04 01:56:46 BST 2003 Olly Betts <olly@survex.com> * HACKING: Bison 1.75 and 1.875a both checked with parsequery.yy - no problems found. Mon Jun 02 20:14:22 BST 2003 Olly Betts <olly@survex.com> * ChangeLog.0: Fixed 2 typos. Mon Jun 02 20:09:01 BST 2003 Olly Betts <olly@survex.com> * NEWS: Updated in preparation for 0.7.0 release. Mon Jun 02 16:25:18 BST 2003 Olly Betts <olly@survex.com> * configure.in: Fix m4 quoting so that MINOR_VERSION is set correctly. Mon Jun 2 15:22:30 2003 James Aylett <james@tartarus.org> * bindings/swig/xapian.i: move QueryParser wrapping to use the new namespace location, and fix include so it can work out of the build directory rather than just the installed library Mon Jun 02 14:15:14 BST 2003 Olly Betts <olly@survex.com> * Makefile.am,configure.in,bindings/Makefile.am,docs/todo.xml: Added --enable-bindings to configure - if this is specified and swig is detected, try to build the language bindings. Mon Jun 02 12:46:18 BST 2003 Olly Betts <olly@survex.com> * backends/database.cc: Fixed parsing of port number in remote stub databases. Sat May 31 15:03:47 BST 2003 James Aylett <james@tartarus.org> * bindings/swig: update so can build Python bindings against latest Xapian; untested, and other languages are likely to fail Fri May 30 02:59:48 BST 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy,extra/queryparsertest.cc: Embedded '.' is now a phrase maker; Ignore `""' in query; Ignore ')' at start of query. Fri May 30 01:22:49 BST 2003 Olly Betts <olly@survex.com> * extra/queryparsertest.cc: Added another 405 queries which fail to parse (courtesy of Arjen van der Meijden). Thu May 29 19:08:28 BST 2003 Olly Betts <olly@survex.com> * docs/todo.xml,extra/queryparsertest.cc: Extracted two more queryparser testcases which need fixing from the old todo list. Thu May 29 17:14:04 BST 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy,extra/queryparsertest.cc: A phrasemaker isn't a phrasemaker unless it's preceded by an alphanumeric, "+", or "-"; If a single term is in (), ignore the (). Thu May 29 16:45:39 BST 2003 Olly Betts <olly@survex.com> * extra/queryparser.yy,extra/queryparsertest.cc: Improved handling of various odd queries encountered in real life applications. Thu May 29 01:50:51 BST 2003 Olly Betts <olly@survex.com> * docs/todo.xml: Updated more. Thu May 29 01:26:50 BST 2003 Olly Betts <olly@survex.com> * api/,backends/database.cc,backends/inmemory/,backends/multi/, backends/muscat36/,backends/net/,backends/quartz/,common/, docs/overview.html,docs/quickstart.html, docs/quickstartexpand.cc.html,extra/queryparser.yy,include/om/om.h, include/xapian/,matcher/,net/,tests/api_db.cc,tests/api_posdb.cc, tests/quartztest.cc,testsuite/: Converted remaining om_xxx types to Xapian::xxx. Wed May 28 23:06:18 BST 2003 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Wed May 28 22:59:17 BST 2003 Olly Betts <olly@survex.com> * tests/api_db.cc: Spelling correction: accross -> across. Wed May 28 20:24:32 BST 2003 Olly Betts <olly@survex.com> * extra/: OmQueryParser -> Xapian::QueryParser, etc. Wed May 28 20:03:48 BST 2003 Olly Betts <olly@survex.com> * api/omdocument.cc: Removed include of defunct header modifieddocument.h. Wed May 28 20:00:56 BST 2003 Olly Betts <olly@survex.com> * api/maptermlist.h,api/omdocument.cc,api/omvalueiterator.cc, backends/inmemory/inmemory_document.cc, backends/inmemory/inmemory_document.h,backends/muscat36/, backends/net/net_document.cc,backends/net/net_document.h, backends/quartz/quartz_document.cc,backends/quartz/quartz_document.h, common/Makefile.am,common/document.h,common/modifieddocument.h, include/xapian/document.h,tests/api_db.cc: Fixed bug in refactored Xapian::Document when modifying a copy. Added regression test. Wed May 28 19:23:26 BST 2003 Olly Betts <olly@survex.com> * common/omenquireinternal.h,include/xapian/database.h, include/xapian/enquire.h,include/xapian/query.h: Where possible, declare classes rather than including headers (in public headers). Wed May 28 17:32:14 BST 2003 Olly Betts <olly@survex.com> * include/xapian/document.h: Just declare "class TermIterator;" rather than including <xapian/termiterator.h>. Tue May 27 16:59:34 BST 2003 Olly Betts <olly@survex.com> * common/progserver.h,common/socketserver.h,net/socketserver.cc: Reduced needless header inclusion. Tue May 27 16:53:40 BST 2003 Olly Betts <olly@survex.com> * bin/omprogsrv.cc,tests/api_db.cc: Use variable name "dbs" in preference to "dbgrp". OmDatabaseGroup hasn't existed for years! Tue May 27 16:51:30 BST 2003 Olly Betts <olly@survex.com> * Makefile.am: Fixed typo: "bin" should be "net" in remote_subdirs. Tue May 27 02:43:32 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/Makefile.am: Removed -I referencing testsuite directory. Tue May 27 02:39:43 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/,docs/tests.html,tests/: Moved quartztest and btreetest to tests/ directory. Tue May 27 02:03:07 BST 2003 Olly Betts <olly@survex.com> * Makefile.am,configure.in,xapian.spec.in,backends/quartz/,bin/, docs/remote.html,netprogs/,tests/api_db.cc,tests/remotetest.cc, testsuite/backendmanager.cc,testsuite/testsuite.cc: Sorted out the source tree structure so that programs which aren't tests which were previously built by "make check" are now be built by "make". Tue May 27 01:01:40 BST 2003 Olly Betts <olly@survex.com> * autoconf/dir_contents,backends/dir_contents, backends/inmemory/dir_contents,backends/multi/dir_contents, backends/muscat36/dir_contents,backends/net/dir_contents, backends/quartz/dir_contents,bindings/dir_contents, common/dir_contents,docs/dir_contents,include/om/dir_contents, include/xapian/dir_contents,include/dir_contents, indexer/dir_contents,indexer/indexgraph/dir_contents, languages/dir_contents,m4/dir_contents,tests/dir_contents: Updated all the dir_contents files - many were out-of-date. Tue May 27 00:42:54 BST 2003 Olly Betts <olly@survex.com> * Makefile.am: Filter any .la files out of dependency_libs. Tue May 27 00:11:18 BST 2003 Olly Betts <olly@survex.com> * netprogs/.cvsignore,netprogs/Makefile.am,netprogs/nettest.cc, tests/.cvsignore,tests/Makefile.am,tests/remotetest.cc: netprogs/nettest -> tests/remotetest. Mon May 26 23:49:49 BST 2003 Olly Betts <olly@survex.com> * netprogs/nettest.cc: Removed test tcpclient1 - the functionality it tests is also tested by tcpmatch1, and it introduces dependencies on internal headers. Mon May 26 22:00:01 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_table_manager.cc: Give a better error message if asked to open a pre-0.6 Quartz database. Mon May 26 20:43:21 BST 2003 Olly Betts <olly@survex.com> * HACKING,Makefile.am,backends/Makefile.am,docs/mkdoc.pl, net/Makefile.am: Fixed to never leave partial files in place of the expected output if a build is interrupted. Mon May 26 16:50:59 BST 2003 Olly Betts <olly@survex.com> * HACKING: Added note adapted from an old email noting why compile-time options are generally best avoided. Mon May 26 12:03:29 BST 2003 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc,testsuite/textfile_indexer.cc: Corrected odd uses of std::string::erase() (instead of `str = str.erase(n, m)' just use `str.erase(n,m)'). Sun May 25 22:59:44 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated. Sun May 25 22:58:28 BST 2003 Olly Betts <olly@survex.com> * languages/Makefile.am: Suppress "unused" warnings in Snowball generated C code. Sun May 25 21:43:34 BST 2003 Olly Betts <olly@survex.com> * api/omtermlistiterator.cc: Removed iostream.h include left over from debugging. Sun May 25 12:27:11 BST 2003 Olly Betts <olly@survex.com> * docs/,extra/omparsequery.h,extra/parsequery.yy,include/Makefile.am, include/xapian.h,include/om/om.h,tests/,testsuite/backendmanager.cc, testsuite/backendmanager.h,testsuite/testutils.h: Created xapian.h header and changed everything to use it. Sat May 24 23:35:46 BST 2003 Olly Betts <olly@survex.com> * api/,backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_document.h, backends/quartz/quartz_database.cc,backends/quartz/quartztest.cc, common/,include/Makefile.am,include/om/om.h, include/om/omvalueiterator.h,include/xapian/, testsuite/backendmanager.cc: OmValueIterator -> Xapian::ValueIterator etc. Fri May 23 15:32:29 BST 2003 Olly Betts <olly@survex.com> * api/,backends/database.cc,backends/inmemory/,backends/muscat36/, backends/net/net_database.cc,backends/net/net_document.cc, backends/net/net_document.h,backends/quartz/,common/,docs/, include/Makefile.am,include/om/om.h,include/om/omdocument.h, include/om/omvalueiterator.h,include/xapian/,indexer/omindexer.h, matcher/,net/,netprogs/nettest.cc,tests/api_db.cc,tests/api_posdb.cc, tests/internaltest.cc,testsuite/: OmDocument -> Xapian::Document, etc. * include/Makefile.am: Fixed library headers to install in xapian and om subdirectories as they should. Fri May 23 15:31:05 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Consistently use int4 for block number - it really should be unsigned and this is a first step. Mon May 19 13:20:46 BST 2003 Olly Betts <olly@survex.com> * configure.in: Create include/xapian if it doesn't exist so that VPATH builds work. Mon May 19 13:17:16 BST 2003 Olly Betts <olly@survex.com> * Makefile.am: Disabled exported symbol regex for now. Tue May 13 05:08:54 BST 2003 Olly Betts <olly@survex.com> * api/,backends/database.cc,backends/dir_contents,backends/inmemory/, backends/multi/,backends/muscat36/,backends/net/,backends/quartz/, common/,docs/,extra/omparsequery.h,include/,matcher/,net/,netprogs/, tests/: OmDatabase -> Xapian::Database, etc. Tue May 13 05:07:41 BST 2003 Olly Betts <olly@survex.com> * NEWS: Corrected spelling: "maintainance" to "maintenance". Tue May 13 00:40:47 BST 2003 Olly Betts <olly@survex.com> * ChangeLog.0,NEWS,backends/inmemory/inmemory_database.h, backends/muscat36/da_database.h,backends/muscat36/db_database.h, backends/quartz/quartz_postlist.h, indexer/indexgraph/omstopwordnode.cc: Spelling correction: "occurence" should be "occurrence". Mon May 12 22:39:27 BST 2003 Olly Betts <olly@survex.com> * api/ompositionlistiteratorinternal.h, include/om/ompositionlistiterator.h: Removed unused files. Mon May 12 20:51:03 BST 2003 Olly Betts <olly@survex.com> * api/omdatabaseinternal.cc,backends/database.cc: Moved database factory functions from api to backends. Fri May 09 04:02:29 BST 2003 Olly Betts <olly@survex.com> * common/expandweight.h: Removed cruft. Fri May 09 03:50:36 BST 2003 Olly Betts <olly@survex.com> * common/rset.h: Removed unused RSetI ctor variant. Fri May 09 02:00:03 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc,common/,docs/overview.html,include/om/om.h, include/xapian/enquire.h,include/xapian/output.h,matcher/, net/socketclient.cc,net/socketcommon.cc,net/socketserver.cc, netprogs/nettest.cc,tests/api_db.cc: OmRSet -> Xapian::RSet. Fri May 09 01:06:32 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc: Fixed MSet::Internal::get_description() to include class name. Fri May 09 01:04:12 BST 2003 Olly Betts <olly@survex.com> * common/rset.h,matcher/expand.cc,matcher/rset.cc: RSetI now uses set<Xapian::docid> (as OmRSet::Internal) rather than vector<RSetItem>. Fri May 09 00:41:20 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc,common/,docs/quickstartexpand.cc.html,matcher/: RSet internal class renamed to RSetI to aid upcoming OmRSet -> Xapian::RSet change. RSetI looks like it can be eliminated later. Thu May 08 23:54:44 BST 2003 Olly Betts <olly@survex.com> * tests/: Removed make_dbgrp() helper function from apitest as it makes the test code less clear without making it easier to write. Thu May 08 01:24:53 BST 2003 Olly Betts <olly@survex.com> * api/ompostlistiterator.cc,docs/overview.html,tests/api_db.cc: Fixed OmWeight and OmMatchDecider in comments, docs, and testsuite. Thu May 08 01:16:30 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc,backends/multi/multi_postlist.h,common/, include/om/om.h,include/xapian/enquire.h,matcher/, net/socketclient.cc,net/socketserver.cc: OmMatchDecider -> Xapian::MatchDecider; OmWeight -> Xapian::Weight; BoolWeight, TradWeight, BM25Weight -> Xapian::. Wed May 07 23:06:47 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btreetest.cc,backends/quartz/quartztest.cc, configure.in,netprogs/nettest.cc,tests/,testsuite/: Minor overhaul of testsuite. Wed May 07 17:54:34 BST 2003 Olly Betts <olly@survex.com> * backends/,common/document.h,matcher/Makefile.am,matcher/document.cc: Removed do_ wrappers for Document methods get_value(), get_all_values() and get_data(). The wrappers used to perform thread locking, but now they're just pointless overhead. Wed May 07 17:24:00 BST 2003 Olly Betts <olly@survex.com> * HACKING: Started section of API structure. Wed May 07 02:16:03 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc,common/expand.h,common/omenquireinternal.h, common/omstringstream.h,docs/quickstart.html,include/om/om.h, include/xapian/enquire.h,include/xapian/output.h,matcher/expand.cc, matcher/multimatch.cc,net/socketcommon.cc,netprogs/nettest.cc, tests/api_db.cc,testsuite/testutils.cc: OmMSet -> Xapian::MSet, similarly for OmMSetIterator, OmESet, and OmESetIterator. Tue May 06 19:31:42 BST 2003 Olly Betts <olly@survex.com> * include/xapian/query.h: New file I failed to checkin on Friday. Tue May 06 14:16:41 BST 2003 Olly Betts <olly@survex.com> * HACKING,configure.in,tests/.cvsignore,tests/Makefile.am, tests/runtest.in,testsuite/testsuite.cc: Sorted out autodetection of valgrind for running testsuite. * HACKING: Added note about how to run gdb on dynamically linked binaries built using libtool. * HACKING: Recommend libtool 1.5 as in 1.4.2 libtool --mode=execute doesn't seem to handle programs with arguments. Tue May 06 00:25:03 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated. Fri May 02 16:58:15 BST 2003 Olly Betts <olly@survex.com> * configure.in: Improved check that LEX is flex - it will now accept "flex" with a path and/or prefix. Fri May 02 16:12:46 BST 2003 Olly Betts <olly@survex.com> * api/,common/,include/,matcher/,net/,netprogs/nettest.cc, netprogs/omtcpsrv.cc,tests/api_db.cc,tests/api_posdb.cc, testsuite/testutils.cc,testsuite/testutils.h: OmEnquire -> Xapian::Enquire, OmQuery -> Xapian::Query, etc. Fri May 02 03:14:28 BST 2003 Olly Betts <olly@survex.com> * tests/internaltest.cc: Reenabled tests for leak checking and fixed them up to work with valgrind. Fri May 02 02:14:28 BST 2003 Olly Betts <olly@survex.com> * testsuite/testsuite.cc,testsuite/testsuite.h: If a testsuite program is run using "valgrind --logfile-fd=255 ./footest" then the harness will show valgrind's output for tests which fail because valgrind spotted errors or leaks (untested under valgrind). Thu May 01 21:25:01 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc,api/omquery.cc,api/omqueryinternal.cc,common/, docs/overview.html,docs/quickstart.html,extra/omparsequery.h, extra/parsequery.yy,extra/parsequerytest.cc,include/,matcher/, net/,netprogs/nettest.cc,tests/api_db.cc,tests/api_nodb.cc, tests/api_posdb.cc: OmQuery -> Xapian::Query, etc. For now I've had to expose Xapian::Query::Internal in xapian/query.h which isn't ideal. Wed Apr 30 20:37:27 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc,common/omenquireinternal.h,include/: typedefs om_* -> Xapian::*, but not changed all internal uses yet. Wed Apr 30 03:51:52 BST 2003 Olly Betts <olly@survex.com> * api/,backends/,common/,extra/omparsequery.h,extra/parsequery.yy, include/,matcher/localmatch.h,matcher/orpostlist.h,tests/api_db.cc: OmPostListIterator -> Xapian::PostListIterator, etc. Wed Apr 30 02:13:31 BST 2003 Olly Betts <olly@survex.com> * docs/matcherdesign.html,include/om/omenquire.h: Merged docs discussing OmMSetIterator::get_collapse_count() into the doxygen documentation comments - they're docs aimed at the API user so really don't belong in the internal matcher design docs. Wed Apr 30 01:08:39 BST 2003 Olly Betts <olly@survex.com> * .cvsignore: Added config.guess. Tue Apr 29 21:08:40 BST 2003 Olly Betts <olly@survex.com> * include/om/om.h,include/xapian/base.h,include/xapian/error.h, include/xapian/errorhandler.h,include/xapian/expanddecider.h, include/xapian/positionlistiterator.h,include/xapian/stem.h, include/xapian/termiterator.h: OM_HGUARD_* -> XAPIAN_INCLUDED_*. Tue Apr 29 20:37:01 BST 2003 Olly Betts <olly@survex.com> * api/,backends/,common/,include/,indexer/omnodeinstanceiterator.h, indexer/ompaditerator.h,net/readquery.h,testsuite/indexer.h: om/omoutput.h -> xapian/output.h; om/omtypes.h -> xapian/types.h. Tue Apr 29 18:28:27 BST 2003 Olly Betts <olly@survex.com> * api/,backends/,common/,include/,matcher/,net/socketserver.cc,tests/, testsuite/backendmanager.cc,testsuite/testsuite.cc: OmTermIterator -> Xapian::TermIterator, etc. Tue Apr 29 17:39:34 BST 2003 Olly Betts <olly@survex.com> * common/omdebug.cc: Setting XAPIAN_DEBUG_LOG caused an exception (with GCC 3.0 at least) unless the value contained %% - fixed. Tue Apr 29 14:29:36 BST 2003 Olly Betts <olly@survex.com> * api/ompositionlistiterator.cc: Fixed to compile in debug mode. Tue Apr 29 01:06:02 BST 2003 Olly Betts <olly@survex.com> * xapian.spec.in: Removed %changelog - it hasn't been reliably updated and only really makes sense when the packaging is done by a third party anyway. Tue Apr 29 01:03:20 BST 2003 Fabrice Colin * xapian.spec.in: Fixed packaging of docs. Tue Apr 29 00:33:21 BST 2003 Olly Betts <olly@survex.com> * HACKING,testsuite/testsuite.cc,testsuite/testsuite.h: valgrind is now mostly hooked into the testsuite. * HACKING: Bison 1.875 doesn't work but Bison 1.875a probably does; suggest CXXFLAGS=-Werror, not CFLAGS. Tue Apr 29 00:14:00 BST 2003 Olly Betts <olly@survex.com> * INSTALL: Updated. Wed Apr 23 14:31:51 BST 2003 Olly Betts <olly@survex.com> * api/omtermlistiterator.cc,api/omtermlistiteratorinternal.h, api/omvalueiteratorinternal.h,backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_positionlist.cc, common/inmemory_positionlist.h: Added InMemoryPositionList ctor which takes the positions data to avoid needing to construct and immediately call set_data(). Wed Apr 23 14:14:34 BST 2003 Olly Betts <olly@survex.com> * api/,backends/,common/,include/,matcher/: OmPositionListIterator -> Xapian::PositionListIterator, etc. Wed Apr 23 13:47:49 BST 2003 Olly Betts <olly@survex.com> * xapian.spec.in: Minor tweaks. Sun Apr 20 22:55:04 BST 2003 Olly Betts <olly@survex.com> * extra/omparsequery.h,include/xapian/stem.h,tests/api_db.cc, tests/api_nodb.cc,tests/api_posdb.cc,testsuite/backendmanager.cc: OmStem -> Xapian::Stem; OmExpandDecider -> Xapian::ExpandDecider. Sun Apr 20 22:52:42 BST 2003 Olly Betts <olly@survex.com> * configure.in,api/omenquire.cc,common/expand.h, common/omenquireinternal.h,docs/overview.html,include/, matcher/expand.cc,tests/api_db.cc: OmExpandDecider -> Xapian::ExpandDecider, etc. Sun Apr 20 22:42:32 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS,api/,backends/,common/,docs/,include/,matcher/,net/, netprogs/,tests/,testsuite/: OmError -> Xapian::Error, etc. Fri Apr 18 22:40:31 BST 2003 Sam Liddicott <sam@liddicott.com> * matcher/multimatch.cc: Fixed so that it updates the OmMSetItem in the MSet with the collapse_count from the OmMSetItem twin in collapse_tab or we lose collapse_count if newer collapsing hits are less relevant than the hits they collapse over. I think we need collapse_tab maybe to keep a reference to the item in the mset? But this works for now. * xapian.spec.in: Fixed to also install include/xapian stuff Fri Apr 18 19:17:56 BST 2003 Olly Betts <olly@survex.com> * configure.in,include/Makefile.am,include/om/.cvsignore, include/om/Makefile.am,include/om/om.h,include/om/om.h.in, include/xapian/.cvsignore,include/xapian/Makefile.am, include/xapian/dir_contents,include/xapian/version.h.in: C++ ABI version checking is now done by xapian/version.h. Fri Apr 18 17:57:21 BST 2003 Olly Betts <olly@survex.com> * configure.in,api/omstem.cc,include/Makefile.am,include/om/om.h.in, include/om/omstem.h,include/xapian/.cvsignore, include/xapian/Makefile.am,include/xapian/base.h, include/xapian/stem.h: Reworked OmStem to use reference counted internals; renamed OmStem to Xapian::Stem and added a #define for compatibility with existing code. Fri Apr 18 17:02:56 BST 2003 Olly Betts <olly@survex.com> * api/,backends/,common/,docs/overview.html,docs/quickstart.html, docs/quickstartexpand.cc.html,extra/,include/om/,matcher/,net/, tests/,testsuite/: Use std::string instead of typedef-ing it as om_termname. The typedef doesn't really buy us anything. Keep a typedef for compatibility with existing code for now. Fri Apr 18 16:12:32 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: GCC 3.2.2 on Redhat 7.1 works with CXXFLAGS=-Werror Thu Apr 10 20:05:27 BST 2003 Olly Betts <olly@survex.com> * common/Makefile.am,common/indexer.h,testsuite/Makefile.am, testsuite/indexer.h: Moved indexer.h from common to testsuite and altered it to use iosfwd instead of iostream and to declare OmDocument as a class instead of include omdocument.h. Thu Apr 10 18:55:55 BST 2003 Olly Betts <olly@survex.com> * xapian-config.nodep.in: Removed --prefix and --exec-prefix - you can't reliably install Xapian with a different prefix to the one it was configured with, yet these options give the impression you can. Thu Apr 10 18:53:53 BST 2003 Olly Betts <olly@survex.com> * .cvsignore: Added config.sub. Thu Apr 10 02:41:50 BST 2003 Olly Betts <olly@survex.com> * NEWS,configure.in: Updated for 0.6.5 release. Thu Apr 10 02:29:12 BST 2003 Olly Betts <olly@survex.com> * docs/doxygen_api_footer.html_tmpl,docs/doxygen_full_footer.html_tmpl, docs/overview.html: Use http://www.doxygen.org/ as URL for doxygen. * docs/overview.html: Fixed bad link to our own website! Thu Apr 10 02:12:00 BST 2003 Olly Betts <olly@survex.com> * Makefile.am,backends/Makefile.am,backends/net/.cvsignore, backends/net/Makefile.am,netprogs/.cvsignore: Renamed libnetdb.la to libremote.la, and LIB_REMOTE1 to LIBNET_LA. Thu Apr 10 02:08:24 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated mingw notes to reflect the change to automatically disable the remote backend when fork() isn't available. Thu Apr 10 01:56:34 BST 2003 Olly Betts <olly@survex.com> * docs/mkdoc.pl: Fixed to only look at directories listed in DIST_SUBDIRS. Thu Apr 10 00:23:28 BST 2003 Olly Betts <olly@survex.com> * configure.in: Automatically disable the remote backend if we don't have fork() since the remote backend requires it in several places; AM_CONDITIONAL ENABLE_SHARED isn't used anywhere so commented it out. Wed Apr 09 22:48:20 BST 2003 Olly Betts <olly@survex.com> * NEWS: Updated with recent changes. Wed Apr 09 22:43:31 BST 2003 Olly Betts <olly@survex.com> * docs/remote_protocol.html: Talk about Server/Client rather than ProgServer/ProgClient; reworked first paragraph. Wed Apr 09 13:23:05 BST 2003 Olly Betts <olly@survex.com> * docs/Makefile.am: Install docs in /usr/share/doc/xapian-core to be FHS compliant. * xapian.spec.in: Package quartzcheck and quartzcompact too; sorted out docs so they all install in the right place. Wed Apr 09 01:16:11 BST 2003 Olly Betts <olly@survex.com> * configure.in,backends/quartz/quartz_log.cc, backends/quartz/quartz_table_manager.cc,common/omdebug.cc, common/omdebug.h: Don't use HAVE_GETPID - if it's not set we assume GetCurrentProcessId() works, so it's cleaner to test with #ifdef WIN32 instead. Tue Apr 08 19:56:57 BST 2003 Olly Betts <olly@survex.com> * PLATFORMS: Updated with the results of many test builds. Tue Apr 08 19:55:15 BST 2003 Olly Betts <olly@survex.com> * configure.in,backends/quartz/Makefile.am,backends/quartz/btree.cc, netprogs/nettest.cc,common/getopt.h,common/omtime.h,common/utils.cc, common/utils.h,tests/Makefile.am,tests/api_db.cc: Now builds with Linux to mingw cross-compiler. * tests/Makefile.am: don't include findheaders.pl in the tarball - it's no longer used. * tests/internaltest.cc: removed "#include <dlfcn.h>" and disabled code which used to use it. * testsuite/backendmanager.cc,testsuite/backendmanager.h: fixed to work if any backends are disabled. Tue Apr 08 16:09:11 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/quartz_log.cc, backends/quartz/quartz_table_manager.cc,testsuite/testsuite.cc: Some fixes for building with Linux to mingw cross-compiler. Tue Apr 08 02:26:51 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/btreetest.cc: If we don't have <sstream> and so are using our own stringstream implementation, we can't pass it as an ostream reference so BtreeCheck::check() can't be run in btreetest. This is only an issue on old compilers (for GCC <= 2.95.2 it seems) so just disable the checking in that case - it's probably not worth the effort of trying to make our stringstream a subclass of ostream. Tue Apr 08 01:20:36 BST 2003 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Updated unfinished hooks to use valgrind in the test harness. Fri Apr 04 03:12:27 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc,api/omstem.cc,common/omdebug.cc: Fixed to compile when configure-d with --enable-debug-verbose --enable-debug=full. Fri Apr 04 03:10:53 BST 2003 Olly Betts <olly@survex.com> * common/omstringstream.h: Removed unused code for writing vector<string> to our own implementation of om_ostringstream. Fri Apr 04 03:08:35 BST 2003 Olly Betts <olly@survex.com> * api/ompostlistiterator.cc: Use "using namespace std;"; replaced gratuitous use of om_ostringstream. Fri Apr 04 03:07:05 BST 2003 Olly Betts <olly@survex.com> * tests/api_nodb.cc: Use "using namespace std;"; removed stray cout debug line. Wed Apr 02 16:54:50 BST 2003 Olly Betts <olly@survex.com> * tests/internaltest.cc: More Sun C++ fixes. * PLATFORMS: Updated. Wed Apr 02 16:39:48 BST 2003 Olly Betts <olly@survex.com> * tests/api_db.cc,net/tcpserver.cc,common/tcpserver.h: More Sun C++ fettling. Wed Apr 02 15:42:54 BST 2003 Olly Betts <olly@survex.com> * matcher/emptymatch.h,tests/api_db.cc: More fixes for Sun C++. Wed Apr 02 15:13:58 BST 2003 Olly Betts <olly@survex.com> * api/omenquire.cc,api/omqueryinternal.cc, backends/inmemory/inmemory_database.h,common/document.h, common/netclient.h,common/netserver.h,common/networkstats.h, common/omlinebuf.h,common/stats.h,include/om/omerror.h, include/om/omerrorhandler.h,indexer/indexgraph/nodetest.cc, indexer/indexgraph/omindexerbuilder.cc, indexer/indexgraph/omindexerinternal.h, indexer/indexgraph/omstopwordnode.cc, indexer/indexgraph/regexcommon.h,indexer/indexgraph/toposort.h, tests/api_db.cc,tests/api_nodb.cc,testsuite/backendmanager.h: More fixes for Sun C++; removed many superfluous ";" after "}". Wed Apr 02 03:45:42 BST 2003 Olly Betts <olly@survex.com> * api/omdocument.cc,api/omenquire.cc,api/omtermlistiteratorinternal.h, backends/inmemory/inmemory_database.h,backends/quartz/btree.cc, backends/quartz/btree.h,common/alltermslist.h,common/expandweight.h, common/getopt.h,common/utils.h,include/om/omenquire.h, matcher/expandweight.cc,matcher/multimatch.cc, testsuite/backendmanager.h: Resolved various issues so that libxapian now builds with Sun's C++ compiler. However libomqueryparser and the test programs don't yet. Tue Apr 01 02:52:57 BST 2003 Olly Betts <olly@survex.com> * backends/quartz/quartz_types.h: use "unsigned int" rather than "unsigned long int" so that quartz_tablesize_t matches om_doccount on 64 bit machines. Mon Mar 31 19:34:54 BST 2003 Olly Betts <olly@survex.com> * AUTHORS,xapian.spec.in: Merged in changes to RPM packaging from Fabrice Colin and reworked further. Fri Mar 28 20:04:04 GMT 2003 Olly Betts <olly@survex.com> * NEWS: Updated. Fri Mar 28 19:53:59 GMT 2003 Olly Betts <olly@survex.com> * README: Removed out-of-date sourceforge references; tidied up wording in places. Fri Mar 28 19:52:11 GMT 2003 Olly Betts <olly@survex.com> * HACKING: Removed out-of-date sourceforge references; tidied up wording in places; docs/tests.txt is now docs/tests.html . Fri Mar 28 17:35:15 GMT 2003 Olly Betts <olly@survex.com> * NEWS: Wrote most of the entries for the next release. Fri Mar 28 17:34:57 GMT 2003 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Thu Mar 27 05:42:41 GMT 2003 Olly Betts <olly@survex.com> * .cvsignore,Makefile.am,configure.in,docs/.cvsignore,docs/Makefile.am, docs/index.html,docs/todo2html.pl: removed machinery to generate TODO, TODO.release, docs/todo.html, and docs/todo-release.html from docs/todo.xml - bug and todo items will be tracked in bugzilla instead. Thu Mar 27 05:25:12 GMT 2003 Olly Betts <olly@survex.com> * docs/todo.xml: Removed completed items and those already fed into bugzilla. Thu Mar 27 05:17:05 GMT 2003 Olly Betts <olly@survex.com> * api/omstem.cc,include/om/omstem.h: Added default OmStem ctor, and "none" language. Both of these give a stemmer object which leaves terms unchanged which should allow for simpler logic in programs using Xapian. The default ctor also removes the need to mess with pointers in some cases. Thu Mar 27 04:52:27 GMT 2003 Olly Betts <olly@survex.com> * NEWS: Started to prepare for next release. Thu Mar 27 04:49:52 GMT 2003 Olly Betts <olly@survex.com> * include/om/omenquire.h,include/om/omerror.h: Improved documentation comments. Thu Mar 27 01:09:04 GMT 2003 Olly Betts <olly@survex.com> * docs/todo.xml,tests/api_db.cc: Added regression test for Quartz bug which caused problems with long terms on machines with signed chars. Wed Mar 26 20:30:18 GMT 2003 Olly Betts <olly@survex.com> * matcher/multimatch.cc: fixed interaction of collapsing and sort_bands == 1. Wed Mar 26 01:04:30 GMT 2003 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fixed sort_bands == 1 bug which would incorrectly reject many documents with a low score. Thu Mar 06 13:04:06 GMT 2003 Sam Liddicott <sam.liddicott@orange.co.uk> * docs/matcherdesign.html: Add writeup about document collapsing. Thu Mar 06 11:32:05 GMT 2003 Sam Liddicott <sam.liddicott@orange.co.uk> * Add collapse_count threshold checking to ensure that most relevant collapsed-away hit is more relevant than threshold (if specified) Mon Mar 03 14:04:56 GMT 2003 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Corrected 2kB to 2KB. Fri Feb 28 10:11:04 GMT 2003 Sam Liddicott <sam.liddicott@orange.co.uk> * Add get_collapse_count() to OmMSetIterator * Add collapse_count to OmMSetItem * Add collapse_count tracking to get_mset() Thu Feb 27 15:28:08 GMT 2003 Olly Betts <olly@survex.com> * common/match.h,common/multimatch.h,matcher/multimatch.cc: Optimise the common sort_bands == 1 case better. * tests/api_db.cc: sortbands2 now tests 1 sortband as well as 10. Wed Feb 26 09:58:01 GMT 2003 Sam Liddicott <sam.liddicott@orange.co.uk> * bootstrap: Fix success message when building in non-src dir as configure is written to the src dir. Mon Feb 24 10:33:52 GMT 2003 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Disabled (for now) sort_bands optimisation which is misbehaving. Fixed forward ordered boolean optimisation broken by last check-in. * tests/api_db.cc: Added test_sortbands2 based on James' test code. Sun Feb 23 20:28:47 GMT 2003 Olly Betts <olly@survex.com> * matcher/multimatch.cc: Fixed problem with sort_bands when asking for an m-set which didn't start with the first match. Sun Jan 05 22:20:33 GMT 2003 Olly Betts <olly@survex.com> * testsuite/index_utils.cc: Handling of ^x was just downright wrong due to a typo. Sun Jan 05 22:19:56 GMT 2003 Olly Betts <olly@survex.com> * backends/quartz/.cvsignore: Ignore quartzcompact binary. Sun Jan 05 22:18:57 GMT 2003 Olly Betts <olly@survex.com> * extra/parsequerytest.cc: Fixed to build with GCC 2.95. Tue Dec 24 20:21:03 GMT 2002 Olly Betts <olly@survex.com> * NEWS: Added omega 0.6.4 changes. Tue Dec 24 19:53:10 GMT 2002 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.in: Version 0.6.4. Tue Dec 24 19:29:27 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc,backends/quartz/quartz_table.cc: Don't bother checking is position_list is empty - just delete it and let the layer below handle not having anything to delete. * backends/quartz/quartz_termlist.cc,docs/todo.xml: Fixed unpacking of termlist on platforms where char is signed. Tue Dec 24 05:48:06 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Fixed double setting of position list when updating a document with term position information. The behaviour before was correct, just inefficient. Mon Dec 23 04:44:46 GMT 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Mon Dec 23 03:22:19 GMT 2002 Olly Betts <olly@survex.com> * extra/parsequery.yy: If a stemmed form with a "." is in the query, include the "." on the form in the unstem multimap. Sun Dec 22 21:44:06 GMT 2002 Olly Betts <olly@survex.com> * extra/omparsequery.h,extra/parsequery.yy,extra/parsequerytest.cc: Added support for searching probabilistic fields (using <field>:<term>). Sat Dec 21 01:38:59 GMT 2002 Olly Betts <olly@survex.com> * configure.in: Don't include "om.h" in APIDOC_SRC, otherwise people building from source will need to regenerate the source docs which will require them to have doxygen installed. Fri Dec 20 15:39:51 GMT 2002 James Aylett <james@tartarus.org> * Bindings build using SWIG 1.3.14u-20020706-1222, with very basic functionality tested with the Python target. Fri Dec 20 12:38:26 GMT 2002 James Aylett <tartarus@users.sourceforge.net> * Tidied up some bindings-related issues, including changes to the Python bindings to work with Python 2.2. This probably won't compile as it stands, but it's disabled anyway. Sun Dec 15 18:23:42 GMT 2002 Olly Betts <olly@survex.com> * api/omdatabase.cc,backends/muscat36/da_database.cc, backends/muscat36/db_database.cc,matcher/phrasepostlist.cc: Muscat 3.6 DA and DB databases don't have word position information. Instead of throwing an exception when this information is requested, return an empty position list (i.e. behave as a quartz database with no position information would). Sun Dec 15 03:08:28 GMT 2002 Olly Betts <olly@survex.com> * backends/muscat36/: DADatabase and DBDatabase ctors need to public so that they can called from the database factory functions. * api/omenquire.cc,backends/muscat36/,backends/quartz/btree.cc, tests/api_db.cc: Fixed compilation warnings. Sat Dec 14 23:02:55 GMT 2002 Olly Betts <olly@survex.com> * docs/quartzdesign.html,docs/todo.xml: Updated. Sat Dec 14 22:57:41 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/Makefile.am,backends/quartz/quartzcheck.cc, backends/quartz/quartzcompact.cc,backends/quartz/quartzdump.cc: Enhanced quartzcheck to check all the btrees in a quartz database if passed a directory name. It also default to "v" rather than "+" which is more appropriate for an end user checking if tables in a non-trivial sized database are corrupt. Added quartzcompact utility to make a copy of a quartz database with full compaction turned on - this results in a smaller database which is faster to search. The next update will result in a lot of block splitting though (since all blocks are as full as possible). Sat Dec 14 16:28:11 GMT 2002 Olly Betts <olly@survex.com> * NEWS: Updated for 0.6.3 release. Sat Dec 14 04:28:16 GMT 2002 Olly Betts <olly@survex.com> * PLATFORMS,configure.in: Version 0.6.3. Sat Dec 14 03:54:08 GMT 2002 Olly Betts <olly@survex.com> * extra/omparsequery.h,extra/parsequery.yy: Create an "unstem" multimap so users can convert the stemmed terms back into their query representation. Sat Dec 14 03:48:27 GMT 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Sat Dec 14 02:37:41 GMT 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Sat Dec 14 02:26:29 GMT 2002 Olly Betts <olly@survex.com> * docs/remote_protocol.html,net/socketclient.cc,net/socketserver.cc: Updated remote protocol description. Sat Dec 14 02:05:59 GMT 2002 Olly Betts <olly@survex.com> * docs/: Updated. Fri Dec 13 23:43:23 GMT 2002 Olly Betts <olly@survex.com> * include/om/om.h.in: Removed bogus // in the middle of a comment. Fri Dec 13 22:17:41 GMT 2002 Olly Betts <olly@survex.com> * configure.in,testsuite/testsuite.cc: Added hooks for using valgrind to find leaks in the test suite. Just need to sort out suitable hooks in valgrind now! Fri Dec 13 19:57:23 GMT 2002 Olly Betts <olly@survex.com> * docs/: OmSettings removal updates. Fri Dec 13 19:44:23 GMT 2002 Olly Betts <olly@survex.com> * bindings/: Made a start updating bindings for OmSettings removal. Fri Dec 13 19:13:36 GMT 2002 Olly Betts <olly@survex.com> * api/omenquire.cc,common/omenquireinternal.h,include/om/omenquire.h, net/socketclient.cc,tests/api_db.cc: Finished removal of OmSettings - all tests now pass once more! Fri Dec 13 16:35:12 GMT 2002 Olly Betts <olly@survex.com> * api/,backends/quartz/quartz_table_manager.h,common/,docs/,include/, matcher/,net/,tests/internaltest.cc: Removed last use of OmSettings. Just a check-point check in - the code compiles, but doesn't pass tests yet. Fri Dec 13 12:53:01 GMT 2002 Olly Betts <olly@survex.com> * netprogs/nettest.cc: Updated disabled code to use new database factory functions; Added "using namespace std;". Thu Dec 12 23:52:32 GMT 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Tidied my desk and converted lots of scraps of paper to todo entries. Thu Dec 12 16:43:22 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Added workaround for shared_level problem; Improved error reporting. Thu Dec 12 01:05:00 GMT 2002 Olly Betts <olly@survex.com> * docs/overview.html: Updated docs on stub databases. Wed Dec 11 20:49:11 GMT 2002 Olly Betts <olly@survex.com> * api/omdatabaseinternal.cc,include/om/omdatabase.h,tests/api_db.cc: Reimplemented stub databases in the new scheme of things. * backends/inmemory/: fully disabled inmemory_errornext and inmemory_abortnext code. Tue Dec 10 13:47:44 GMT 2002 Olly Betts <olly@survex.com> * HACKING: Added note that Bison 1.50 seems to work with Xapian's .yy files. Mon Dec 09 20:13:00 GMT 2002 Olly Betts <olly@survex.com> * api/omenquire.cc,common/omenquireinternal.h,docs/todo.xml, include/om/omenquire.h,tests/api_db.cc: OmEnquire::get_eset() now takes a flags argument of bit constants |-ed together instead of 2 bools. Mon Dec 09 12:52:38 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: Applied patch from Martin Porter with better fix for sequential addition bug. Mon Dec 09 09:57:42 GMT 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Mon Dec 09 07:32:46 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartztest.cc: Corrected name of database used by test adddoc2 (was using testdb_adddoc1!) Mon Dec 09 05:10:09 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc,backends/quartz/quartz_table.cc: More fettling. Mon Dec 09 04:48:39 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/: Minor code fettle. Mon Dec 09 03:32:11 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: Gratuitous layout fettling. Mon Dec 09 03:30:27 GMT 2002 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Improved wording and punctuation in 3 places. Mon Dec 09 03:25:09 GMT 2002 Olly Betts <olly@survex.com> * common/,docs/todo.xml,include/om/omenquire.h,matcher/, net/socketclient.cc,net/socketserver.cc,netprogs/omprogsrv.cc, netprogs/omtcpsrv.cc: Fixed the remote backend to handle non-default weighting schemes. You can now even implement your own weighting scheme and use it with the remote backend provided you register it with SocketServer at runtime. Sat Dec 07 21:18:39 GMT 2002 Olly Betts <olly@survex.com> * PLATFORMS: Updated. Sat Dec 07 21:09:52 GMT 2002 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.in: Version 0.6.2. Sat Dec 07 20:24:46 GMT 2002 Olly Betts <olly@survex.com> * NEWS: Updated. Sat Dec 07 20:21:41 GMT 2002 Olly Betts <olly@survex.com> * docs/quickstart.html: Fixed parameters passed to OmQuartz__open(). Sat Dec 07 20:17:42 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartztest.cc: Fixed parameters passed to OmQuartz__open(). Sat Dec 07 16:41:25 GMT 2002 Olly Betts <olly@survex.com> * docs/tests.html,testsuite/testsuite.cc: The testsuite won't install its signal handler if XAPIAN_SIG_DFL is set. Sat Dec 07 16:39:01 GMT 2002 Olly Betts <olly@survex.com> * HACKING: OM_DEBUG_XXX -> XAPIAN_DEBUG_YYY. Sat Dec 07 04:48:37 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: max_item_size wasn't being set due to some over-zealous code pruning. It was defaulting to 0, and was causing the code to write off the end of allocated memory blocks. Sat Dec 07 03:22:05 GMT 2002 Olly Betts <olly@survex.com> * backends/inmemory/,matcher/multimatch.cc: Minor code tidying. * backends/quartz/btree.cc: Added Assert in Btree::add() to detect running off end of buffer. Sat Dec 07 03:20:52 GMT 2002 Olly Betts <olly@survex.com> * matcher/localmatch.cc: fixed handling of wtscheme() - we were trying to use it for the extra weights, and then double deleting it! Fri Dec 06 23:05:22 GMT 2002 Olly Betts <olly@survex.com> * api/omstem.cc,backends/quartz/,common/omdebug.cc,common/utils.h, matcher/tradweight.cc,net/socketcommon.cc,net/tcpclient.cc, tests/api_db.cc: Fixed to build with configure --enable-debug=full. Fri Dec 06 23:01:08 GMT 2002 Olly Betts <olly@survex.com> * common/omdebug.cc,common/omdebug.h: Fixed permissions on newly created log file (was getting 000!); Simplified class internals; Renamed env vars: OM_DEBUG_FILE is now XAPIAN_DEBUG_LOG, OM_DEBUG_TYPES is now XAPIAN_DEBUG_FLAGS (old versions still work for now). Fri Dec 06 22:58:57 GMT 2002 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Fixed so running "gdb .libs/apitest" finds srcdir (for an in-tree build at least). Fri Dec 06 03:51:35 GMT 2002 Olly Betts <olly@survex.com> * common/,include/om/omenquire.h,matcher/localmatch.h,matcher/rset.cc, matcher/stats.cc: Fixed to compile with GCC 3.0. Thu Dec 05 23:28:29 GMT 2002 Olly Betts <olly@survex.com> * include/om/omdatabase.h: Added missing "std::". Thu Dec 05 23:27:57 GMT 2002 Olly Betts <olly@survex.com> * docs/remote.html: Updated from OmSettings to factory functions. Thu Dec 05 23:26:42 GMT 2002 Olly Betts <olly@survex.com> * PLATFORMS: ixion is actually Linux 2.2. Thu Dec 05 04:32:40 GMT 2002 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc: Fixed BackendManager::do_getdb_quartz() and do_getwritedb_quartz() to work correctly and so resolved some test failures. Wed Dec 04 03:35:12 GMT 2002 Olly Betts <olly@survex.com> * api/omdatabaseinternal.cc,backends/quartz/,docs/todo.xml, include/om/omdatabase.h: Replace create and allow_overwrite boolean flags with OM_DB_XXX constants. And we now support OM_DB_CREATE_OR_OPEN which is a common action to want to perform, but was fiddly to achieve before. Tue Dec 03 23:59:30 GMT 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Tue Dec 03 23:26:27 GMT 2002 Olly Betts <olly@survex.com> * api/omenquire.cc,backends/quartz/quartz_postlist.h,common/, docs/omsettings,docs/todo.xml,include/om/omenquire.h,matcher/, net/socketcommon.cc,net/socketserver.cc,tests/api_db.cc, tests/api_posdb.cc: Weighting schemes are now specified by passing in a weighting object, rather than via OmSetttings. The weight class can be sub-classed by the library user to allow them to specify their own weighting scheme). Everything works apart from the remote backend where the weighting scheme type and parameters aren't passed across the link. Mon Dec 02 20:12:31 GMT 2002 Olly Betts <olly@survex.com> * NEWS: Updated with changes since 0.6.1. Mon Dec 02 19:44:37 GMT 2002 Olly Betts <olly@survex.com> * api/ompostlistiterator.cc,backends/multi/multi_postlist.h, common/irweight.h,common/leafpostlist.h,matcher/: IRWeight renamed to OmWeight in preparation for making it externally visible. Mon Dec 02 18:10:55 GMT 2002 Olly Betts <olly@survex.com> * matcher/bm25weight.cc,matcher/irweight.cc,matcher/tradweight.cc, matcher/tradweight.h: Started to rework weighting scheme code. Mon Dec 02 17:36:21 GMT 2002 Olly Betts <olly@survex.com> * api/omdatabaseinternal.cc,backends/quartz/quartz_table_manager.cc: Tidying up after the recent backend reworking. Mon Dec 02 03:59:39 GMT 2002 Olly Betts <olly@survex.com> * matcher/: Changed BoolWeight ctor not to take an OmSettings parameter which is simply ignored. Mon Dec 02 01:35:41 GMT 2002 Olly Betts <olly@survex.com> * include/om/omdatabase.h: Added documentation comments for all of the database factory functions. Sun Dec 01 21:45:49 GMT 2002 Olly Betts <olly@survex.com> * include/om: Added \file documentation comments so doxygen extracts documentation for functions too. * docs/: Made a start on updating for the new database factory functions. Sun Dec 01 15:38:10 GMT 2002 Olly Betts <olly@survex.com> * api/,backends/,common/,docs/omsettings,docs/todo.xml, include/om/omdatabase.h,netprogs/,tests/,testsuite/backendmanager.cc, testsuite/backendmanager.h: No longer use OmSettings to specify parameters for constructing databases. Instead there's a factory function for each database type - temporary naming scheme is OmXxx__open(), mostly because it's easy to grep for later. At present stub databases and the machinery in InMemory to allow the multierrhandler1 test aren't working. Everything else should be. Thu Nov 28 20:15:47 GMT 2002 Olly Betts <olly@survex.com> * api/omenquire.cc,common/omenquireinternal.h,docs/omsettings, docs/overview.html,include/om/omenquire.h,tests/api_db.cc: No longer use OmSettings in OmEnquire::get_eset(); fixed reversed sense of use_query_terms (and fixed reversed sense test in apitest which meant this wasn't spotted). Thu Nov 28 20:14:54 GMT 2002 Olly Betts <olly@survex.com> * docs/index.html: Link to annotated class lists in doxygen generated documentation rather than rather empty index page. Thu Nov 28 02:23:33 GMT 2002 Olly Betts <olly@survex.com> * configure.in,NEWS: Version 0.6.1. Thu Nov 28 01:38:05 GMT 2002 Olly Betts <olly@survex.com> * PLATFORMS: Updated. Thu Nov 28 01:33:25 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/: Fixed to compile with GCC 3.0. Thu Nov 28 01:32:31 GMT 2002 Olly Betts <olly@survex.com> * PLATFORMS: Updated with test results from 0.5.4 release. Wed Nov 27 22:11:38 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/btree.h: fixed "public:" bodge added during factoring-out of BtreeCheck. Wed Nov 27 05:01:14 GMT 2002 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.in: Updated for 0.6.0 release. Wed Nov 27 05:00:39 GMT 2002 Olly Betts <olly@survex.com> * AUTHORS,HACKING: Removed or replaced sourceforge.net URLs. Wed Nov 27 04:58:29 GMT 2002 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Reworded "under development" warning; Updated Btree::check(), which is now BtreeCheck::check(). Wed Nov 27 01:36:26 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/btreecheck.cc,backends/quartz/btreecheck.h, backends/quartz/btreetest.cc: Tidied up BtreeCheck - btreetest now sends check output to tout so it's only displayed if the check fails (or btreetest is run with -v). Tue Nov 26 19:47:24 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/: Split the btree checking code out into a separate file, so it's not linked in when we don't need it. Tue Nov 26 05:47:32 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Suppress all output from a successful Btree::Check() is no options are specified, so that btreetest generates clean output when all tests pass. Tue Nov 19 02:08:30 GMT 2002 Olly Betts <olly@survex.com> * NEWS: Updated. Tue Nov 19 01:46:35 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h, backends/quartz/btreetest.cc,backends/quartz/quartztest.cc, docs/quartzdesign.html: Quartz B-tree minimum blocksize is now 2048 bytes (as was in fact documented already). This means the max term length is now always 252 bytes. Mon Nov 18 22:16:57 GMT 2002 Olly Betts <olly@survex.com> * api/omenquire.cc,matcher/expand.cc,matcher/expandweight.cc: Bit of an expand tidy up. Mon Nov 18 19:49:09 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_termlist.cc: I'd put the new termlist stuff in the disabled branch of a "#ifdef" so it wasn't being used! Fixed this and made it actually compile as an encore. Mon Nov 18 02:51:55 GMT 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Retargetted 0.6 tasks for 0.7 or 0.8; updated a few entries. Mon Nov 18 02:28:55 GMT 2002 Olly Betts <olly@survex.com> * docs/quartzdesign.html,backends/quartz/: We can just store the term name raw in position list key, since we know its length from the key length; tweaked storing of deltas to store (delta - 1) as 0 is invalid. Mon Nov 18 00:47:54 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_termlist.cc: First cut of compressed term lists. Sun Nov 17 20:19:23 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_lexicon.cc,backends/quartz/quartz_values.cc: Use new pack_uint_last() and unpack_uint_last() where appropriate. Sun Nov 17 20:03:24 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_record.cc,backends/quartz/quartz_utils.h: Added a more compact integer packing for the last integer in a key or tag. Sun Nov 17 18:54:05 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: more int -> bool changes. * backends/quartz/btree_base.cc: Clarified comments. Sun Nov 17 14:19:30 GMT 2002 Olly Betts <olly@survex.com> * docs/quartzdesign.html: Updated to reflect removal of Btree_item. Sun Nov 17 14:12:12 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/: Eliminate Btree_item - wherever it's used we only actually want the tag or key, so it's cleaner and clearer to just pass a pointer to a string. Sun Nov 17 13:34:08 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/,docs/quartzdesign.html: Quartz mini-overhaul: Removed Btree::Bcursor_create() and just made Bcursor's ctor public - now you can just create a Bcursor on the stack rather than being forced to use AutoPtr; removed valid_handle - the code is clearer without it; sys_close now inlined from header; changed int/char to bool where appropriate; changed some methods which now always throw on error to return void not bool; moved quartztest's test_btree1 to btreetest as test_simple1; changed a few remaining uses of "struct Btree" to just plain "Btree"; brought the btree docs in quartzdesign.html up-to-date, apart from the error handling section. Sat Nov 16 01:45:28 GMT 2002 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.cc, backends/inmemory/inmemory_database.h, backends/quartz/quartz_table.cc: Tidying up odds and ends. Fri Nov 15 17:46:38 GMT 2002 Olly Betts <olly@survex.com> * NEWS: Updated ready for 0.6.0 with changes made so far. Fri Nov 15 17:06:58 GMT 2002 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Avoid double debug output if we get a signal inside the testsuite code itself. Fri Nov 15 17:00:15 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/btree_api.txt,docs/Makefile.am,docs/index.html, docs/quartzdesign.html,docs/remote_protocol.html, net/README_progprotocol.txt: Converted Btree API documentation to HTML and add it to the quartz design document; converted the (seriously out of date) remote protocol document to HTML, moved it into docs/, and linked it in. Fri Nov 15 13:05:06 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_record.cc,backends/quartz/quartz_utils.h, docs/quartzdesign.html: Changed encoding of keys for quartz record and termlist tables. Fri Nov 15 12:54:39 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_table_entries.cc: Fixed QuartzTableEntries::empty() which would never return true before. Fri Nov 15 03:08:49 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/btree_api.txt: Fixed typos. Fri Nov 15 03:06:41 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartztest.cc,backends/quartz/btreetest.cc: added another test (btreetest: emptykey1) and added extra checks to existing tests. Thu Nov 14 01:59:06 GMT 2002 Olly Betts <olly@survex.com> * matcher/multimatch.cc,tests/api_db.cc,docs/todo.xml: match_sort_bands code fettled, and added regression test for the >100% problem (test_sortbands1). Tue Nov 12 00:00:44 GMT 2002 Olly Betts <olly@survex.com> * backends/inmemory/inmemory_database.h,backends/quartz/, include/om/omtypes.h: Pushed average length calc down into QuartzRecordManager; removed unnecessary types om_totlength and om_termid. Mon Nov 11 18:10:23 GMT 2002 Olly Betts <olly@survex.com> * docs/index.html: Fixed broken link. Mon Nov 11 18:07:41 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/,docs/quartzdesign.html: Store next free docid and total doc length in the same tag in QuartzRecord. Wed Nov 06 22:38:40 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc,backends/quartz/quartz_database.h: Eliminated QuartzDatabase::get_doccount_internal() and QuartzDatabase::get_avlength_internal() - now the thread locking code has gone, the indirection serves no purpose. Wed Nov 06 17:50:12 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/: Removed QuartzBufferedTable::write() (unused and unimplemented), and merged QuartzBufferedTable::write_internal() into QuartzBufferedTable::write_internal(). Wed Nov 06 01:56:41 GMT 2002 Olly Betts <olly@survex.com> * backends/quartz/: disable quartz lexicon table unless USE_LEXICON is defined. Wed Nov 06 01:54:30 GMT 2002 Olly Betts <olly@survex.com> * docs/overview.html: Corrected link to API docs on website. Mon Nov 04 02:35:22 GMT 2002 Olly Betts <olly@survex.com> * include/om/omerror.h,net/omerr_string.cc,netprogs/omprogsrv.cc, netprogs/omtcpsrv.cc,testsuite/testsuite.cc: don't use typeid(). Mon Nov 04 01:31:47 GMT 2002 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Catch by const reference. Mon Nov 04 01:18:53 GMT 2002 Olly Betts <olly@survex.com> * HACKING: Added note about use of various C++ features. Also tidied up various odds and ends. Sun Nov 03 22:59:34 GMT 2002 Olly Betts <olly@survex.com> * api/omdatabaseinternal.cc: Eliminated entirely superfluous use of dynamic_cast. Sun Nov 03 22:46:34 GMT 2002 Olly Betts <olly@survex.com> * matcher/multimatch.cc,common/database.h,common/net_database.h: Replace Database::is_network() with Database::as_networkdatabase() which returns a pointer or NULL. This eliminates the need to use dynamic_cast in MultiMatch. Wed Oct 23 05:24:14 BST 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Resolved TODO entry by consulting Stroustrup (it is safe to throw and catch an exception in a destructor, even when that destructor is being called during a stack unwind caused by an exception being thrown). Wed Oct 16 20:05:39 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/btreetest.cc: Suppress (most) output from Btree::check unless verbose (-v) is enabled. Mon Oct 14 15:13:18 BST 2002 Olly Betts <olly@survex.com> * testsuite/testsuite.cc: Fixed compilation problem with std:: on GCC 3.2. Mon Oct 14 15:07:03 BST 2002 Olly Betts <olly@survex.com> * api/omstem.cc: French and Finnish stemmers were switched! Mon Oct 14 02:49:12 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_utils.h: change sort preserving packing for strings. This one does better provided the strings being packed don't contain many zero bytes (which is true of the termnames which we pack with this). Mon Oct 14 02:45:20 BST 2002 Olly Betts <olly@survex.com> * tests/stemtest.cc: Updated to reflect new naming of stemming test data. Sun Oct 13 17:06:26 BST 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Sun Oct 13 16:40:24 BST 2002 Olly Betts <olly@survex.com> * api/omstem.cc,backends/quartz/,common/,extra/omparsequery.h,matcher/, net/socketclient.cc: Added private copy ctors and assignment operators to classes with pointer members (which shouldn't be copied). Sun Oct 13 15:12:40 BST 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Sun Oct 13 02:33:57 BST 2002 Olly Betts <olly@survex.com> * .cvsignore: Added install-sh. Sun Oct 13 02:16:31 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/{quartz_lexicon.cc,quartz_lexicon.h}: key to lexicon is now simply the termname - no need to encode the length there too since the Btree knows how long the key is. * backends/quartz/{quartz_table_manager.cc,quartz_values.cc}: finished the terminology change from "attribute" to "value" (including renaming the btree files, which is why it wasn't done before). Sun Oct 13 02:09:00 BST 2002 Olly Betts <olly@survex.com> * tests/api_db.cc: Snowball stems "this" to "this" rather than "thi", so update apitest source to reflect this. Sun Oct 13 01:03:04 BST 2002 Olly Betts <olly@survex.com> * configure.in: Removed references to languages/*/Makefile. Sun Oct 13 00:50:01 BST 2002 Olly Betts <olly@survex.com> * docs/indexerquickstart.html,docs/quartzdesign.html: Updated. Sat Oct 12 23:34:52 BST 2002 Olly Betts <olly@survex.com> * languages/{api.c,api.h,header.h,utilities.c}: 4 new files. Sat Oct 12 23:29:50 BST 2002 Olly Betts <olly@survex.com> * languages/: removed all the old .cvsignore files. Sat Oct 12 22:34:01 BST 2002 Olly Betts <olly@survex.com> * api/omstem.cc,docs/stemming.html,docs/todo.xml,languages/: Replaced our stemmers with those from Snowball. Note that these give better results, but this also means that existing databases won't work quite correctly if they contain stemmed terms. Sat Oct 12 17:17:26 BST 2002 Olly Betts <olly@survex.com> * configure.in,NEWS: Version 0.5.3. Sat Oct 12 16:49:50 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS,api/omvalueiteratorinternal.h,extra/parsequery.yy, include/om/omdocument.h: Fixed std:: namespace issues to allow compilation with GCC 3 once again. Sat Oct 12 15:43:22 BST 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Fri Oct 11 02:14:47 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/Makefile.am: Include test data for btreetest in distribution tarball. Fri Oct 11 01:43:03 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/{z_note,z_sequence,z_Cversion/,z_make/,z_test/}: Removed all the unused old C btree stuff. Thu Oct 10 17:32:10 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/{btree.cc,btree.h}: More Btree tidying. Thu Oct 10 17:14:28 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/{bcursor.cc,btree.cc,btree.h}: Btree::prev* and Btree::next* are no longer static methods. * backends/quartz/btree.h: Tweaked comments so doxygen will understand them. Thu Oct 10 16:12:29 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_postlist.cc: return docids rather than passing in a pointer to fill in. Thu Oct 10 16:08:22 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/quartzcheck.cc: Fixed #include-d files; Example in usage message now uses the directory where omega looks for its database by default. Thu Oct 10 15:25:10 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/btree.cc,backends/quartz/btree.h: More comment improvements; Removed superfluous Cursor parameter from Btree::add_kt() - it always gets passed the C member of Btree which we have available anyway. Thu Oct 10 14:49:11 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/btree.cc: Removed "struct" from in front of Btree and Cursor; Updated and reformatted many comments. Thu Oct 10 03:45:56 BST 2002 Olly Betts <olly@survex.com> * testsuite/backendmanager.cc: Reworded comment to make it clearer. Thu Oct 10 03:29:52 BST 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Wed Oct 09 14:05:07 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/{quartz_alltermslist.cc,quartz_alltermslist.h, quartz_database.cc}: Implemented QuartzAllTermsList::get_approx_size(). Wed Oct 09 10:46:42 BST 2002 Olly Betts <olly@survex.com> * extra/omparsequery.h,include/om/omvalueiterator.h: Don't use "using std::foo;" in externally visible headers. Wed Oct 09 10:26:05 BST 2002 Olly Betts <olly@survex.com> * extra/omparsequery.h: fixed unused parameter warning. Tue Oct 08 20:32:46 BST 2002 Olly Betts <olly@survex.com> * tests/stemtest.cc: Added missing space to output. Tue Oct 08 20:10:35 BST 2002 Olly Betts <olly@survex.com> * With GCC, add warning flags "-Wall -W" rather than "-Wall -Wunused" (-Wall implies -Wunused anyway). Fixed all the warnings this throws up, except in languages/ (that code is to be replaced with Snowball soon). Tue Oct 08 19:57:03 BST 2002 Olly Betts <olly@survex.com> * testsuite/: Disable colour test output if stdout isn't a terminal; reworked check for broken exception handling as the previous version never seemed to fire; added "using" for all the things we want from std::; improved how signal handlers are set and unset; report exception class for exceptions derived from OmError rather than a blanket "OMEXCEPT"; added private copy ctor and assignment to test_driver to prevent copying. Tue Oct 08 19:53:02 BST 2002 Olly Betts <olly@survex.com> * include/om/Makefile.am: remove include/om/om.h on "make distclean", not "make clean". Otherwise "./configure ; make clean ; make" fails which is wrong. Sun Oct 06 18:37:39 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS: Updated. * docs/: Removed Martin's paper - background information for stemmers is best left to the Snowball documentation. Sat Oct 05 20:31:55 BST 2002 Olly Betts <olly@survex.com> * NEWS,configure.in: 0.5.2 release. Sat Oct 05 03:01:49 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS: Updated. Sat Oct 05 02:09:35 BST 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Fri Oct 04 22:34:12 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS,testsuite/testutils.h: Code to spot mishandled exceptions doesn't always work - noted this in PLATFORMS, and tweaked the code a little. Fri Oct 04 19:08:08 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/.cvsignore: Added quartzcheck. Fri Oct 04 18:24:55 BST 2002 Olly Betts <olly@survex.com> * docs/: converted all text docs to HTML (except omsettings which will has odd markup (LaTeX?) and will probably soon be obsolete anyway). * docs/todo.xml: updated. Fri Oct 04 17:18:33 BST 2002 Olly Betts <olly@survex.com> * net/socketcommon.cc: Fixed handling of timeouts in the past. Fri Oct 04 13:51:39 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS,testsuite/testutils.h: Use typeid() to spot when GCC 2.95 mishandles an exception, and don't count this as a test failure. Fri Oct 04 03:59:29 BST 2002 Olly Betts <olly@survex.com> * api/omdatabaseinternal.cc,backends/multi/multi_termlist.h, matcher/multimatch.cc,matcher/networkmatch.cc,matcher/networkmatch.h: pulled uses of dynamic_cast<> to higher up in the code. Fri Oct 04 03:53:44 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS,backends/quartz/{btree.cc,quartz_metafile.cc}: Fixed quartz problems on platforms where sizeof(long) != 4. * PLATFORMS: Updated in the light of investigations into test failures on x86 Redhat Linux - only multierrhandler1 with the remote backend is actually a problem. Wed Oct 02 16:32:48 BST 2002 Olly Betts <olly@survex.com> * NEWS: Updated for 0.5.1 release. Wed Oct 02 16:01:50 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS,configure.in: 0.5.1 release. Tue Oct 01 13:34:24 BST 2002 Olly Betts <olly@survex.com> * docs/: tweaked navigation links in index.html; converted bm25 text document to HTML and linked it in; added todo entry to locate the "illusion of control" paper. Tue Oct 01 12:48:41 BST 2002 Olly Betts <olly@survex.com> * docs/: renamed intro.html to install.html, and userman.html to overview.html; changed navbar on index.html. Tue Oct 01 12:37:20 BST 2002 Olly Betts <olly@survex.com> * bootstrap: fixed adding of directory with xapian.m4 in to ACLOCAL_FLAGS; added trap to tell user if bootstrapping failed. Tue Oct 01 12:33:38 BST 2002 Olly Betts <olly@survex.com> * HACKING,Makefile.am,bootstrap,buildall,xapian.spec.in, autoconf/dir_contents,docs/todo.xml: Added bootstrap script as a replacement for buildall (buildall left in place for now until bootstrap receives wider testing). Tue Oct 01 12:19:32 BST 2002 Olly Betts <olly@survex.com> * HACKING,Makefile.am,configure.in,PLATFORMS: require automake 1.6.3 (and hence autoconf 2.54) to fix problem building tests/internaltest with Solaris make. Mon Sep 30 20:30:42 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS: Improved wording about bogus OMEXCEPT failures); Added results for OpenBSD 3.0 and Solaris 8 on x86 (both work). Mon Sep 30 20:25:03 BST 2002 Olly Betts <olly@survex.com> * docs/: Pruned .cvsignore; removed cvs.html (link to CVS info on website directly); removed using_stemmers.html - those wishing to use the C API to the stemmers ought to look at Snowball instead; include HTML versions of quickstart*.cc in tarballs; improved wording in various places. Mon Sep 23 19:33:31 BST 2002 Olly Betts <olly@survex.com> * net/Makefile.am: Fixed building of readquery.cc from readquery.ll. Sun Sep 22 03:57:20 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS: Updated. Fri Sep 20 15:09:25 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS: Added note about bogus OMEXCEPT test failures with GCC 2.95. Fri Sep 20 01:36:35 BST 2002 Olly Betts <olly@survex.com> * NEWS,PLATFORMS,configure.in: 0.5.0 release! Fri Sep 20 01:01:52 BST 2002 Olly Betts <olly@survex.com> * Makefile.am: Perl module we need to process todo.xml is XML::Parser not XML, so corrected error message. * configure.in: Commented out MSG_WARN() when Perl modules XML::Parser or Text::Format aren't found. In a release tarball, this only matters if you modify docs/todo.xml, and the warning message(s) may alarm those building the software. Thu Sep 19 00:57:23 BST 2002 Olly Betts <olly@survex.com> * .cvsignore: Added depcomp, missing, mkinstalldirs. Thu Sep 19 00:49:01 BST 2002 Olly Betts <olly@survex.com> * HACKING: Added note about safe way to generate files in make rules. Thu Sep 19 00:45:51 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS: Tested on Solaris 7 with GCC 2.95.3 (works) and Solaris 8 Sun Workshop C++ compiler (fails to compile). Thu Sep 19 00:41:53 BST 2002 Olly Betts <olly@survex.com> * ar-wrapper-solaris,ltconfig,Makefile.am: libtool 1.4 doesn't use ltconfig, and that's the only thing that uses ar-wrapper-solaris, so removed both. Fairly sure the problem they address was fixed in libtool so time ago. Xapian doesn't build with Sun's C++ compiler at the moment anyway, so it's moot for the release. Thu Sep 19 00:26:37 BST 2002 Olly Betts <olly@survex.com> * docs/todo2html.pl: Handle release "*" specially - it indicates tasks to be done for each release. Thu Sep 19 00:05:40 BST 2002 Olly Betts <olly@survex.com> * configure.in,include/om/om.h.in: Tweaked how generation of om.h works to avoid problems with Sun's C++ compiler. Wed Sep 18 19:21:47 BST 2002 Olly Betts <olly@survex.com> * net/Makefile.am,net/readquery.ll: #include <config.h> must be included before any other headers, but that seems to be impossible to arrange in lex/flex so we use echo and cat in the Makefile to arrange this. Wed Sep 18 17:35:00 BST 2002 Olly Betts <olly@survex.com> * include/om/Makefile.am: We want to install om/om.h, but not distribute it or people with a different compiler to whoever ran "make dist" will get the "ABI mismatch" error. Wed Sep 18 13:16:14 BST 2002 Richard Boulton <richard@tartarus.org> * docs/.cvsignore: Add docs/doxygen_full_warnings. * include/om/.cvsignore: Add include/om/om.h. Wed Sep 18 11:06:12 BST 2002 James Aylett <tartarus@users.sourceforge.net> * include/om/Makefile.am: We really should install om/om.h, or applications can't build against us. Wed Sep 18 05:10:24 BST 2002 Richard Boulton <richard@tartarus.org> * docs/doxygen_{api,full}_header.html_tmpl: Remove meta robots tag - why shouldn't this be indexed? Now the omega I set up pointing at copies of this should actually index something. Wed Sep 18 04:52:20 BST 2002 Olly Betts <olly@survex.com> * configure.in: Merged two sed invocations into one - we don't want configure to run any slower than it has to. Wed Sep 18 04:17:58 BST 2002 Richard Boulton <richard@tartarus.org> * configure.in: Don't put confdefs.h into SOURCEDOC_H_SRC - this file only exists during the configure run. Fixes failure of "make doxygen_docs" in docs/ Tue Sep 17 13:07:39 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Updated PLATFORMS file. Mon Sep 16 19:40:02 BST 2002 Olly Betts <olly@survex.com> * docs/todo.xml: Updated. Mon Sep 16 18:09:29 BST 2002 Olly Betts <olly@survex.com> * tests/api_db.cc: don't run consistency1 test on the remote backend - it's particularly slow with that, and testing it there doesn't actually improve the test coverage really. Mon Sep 16 17:15:46 BST 2002 Olly Betts <olly@survex.com> * configure.in,docs/Makefile.am: Added workaround for pattern length limits in Solaris sed. Mon Sep 16 17:13:20 BST 2002 Olly Betts <olly@survex.com> * include/om/om.h.in: Added explanatory comments for developers. Mon Sep 16 14:22:48 BST 2002 James Aylett <tartarus@users.sourceforge.net> * om/om.h needs to be in CLEANFILES or make distcheck fails Mon Sep 16 13:37:04 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Now om/om.h is built from om/om.h.in, it may be in the build directly. Consequently, we need to add that to the include path for VPATH builds. Plus (also for VPATH), om/om.h needs to be output after AC_OUTPUT, because otherwise the destination directory may not exist. Mon Sep 16 04:19:54 BST 2002 Olly Betts <olly@survex.com> * configure.in,docs/todo.xml,include/om/Makefile.am,include/om/om.h, include/om/om.h.in: When building the library with GCC, generate include/om/om.h with preprocessor code to check that any version of GCC used to build applications has a matching C++ ABI. This means that users get a nice explanatory error message rather than a confusing link failure (or worse a program which builds but crashes). Another benefit is that the check happens near the start of compilation of the first source file which uses Xapian in the user's application, rather than during the first attempt to link with Xapian. Sun Sep 15 01:42:19 BST 2002 Olly Betts <olly@survex.com> * net/tcpserver.cc: Solved std mystery - "using namespace std;" was inside a #ifdef which wasn't enabled! Sun Sep 15 01:21:34 BST 2002 Richard Boulton <richard@tartarus.org> * tests/tcpserver.cc: Hacky fix to compile with gcc-3.2 snapshot in Debian unstable - add a FIXME to work out why this is needed. * todo: add entry about checking ABI versions using __GXX_ABI_VERSION macro. Sat Sep 14 22:39:11 BST 2002 Olly Betts <olly@survex.com> * configure.in,tests/Makefile.am,tests/internaltest.cc: It turns out that internaltest *does* need -fno-access-control, so put it back. Not quite sure how I missed this. Changed to skip refcnt tests when -fno-access-control isn't available (rather than omitting them entirely). * tests/api_db.cc: Fixed compile problem. Sat Sep 14 21:44:34 BST 2002 Olly Betts <olly@survex.com> * internaltest: doesn't actually need -fno-access-control these days, so removed it and the configure tests for it. * stemtest: recoded to be all in C++ and removed use of intermediate files for random data tests - now runs ~15% faster; no longer need HAVE_PERL so removed from configure. * apitest: use C++ streams rather than stdio; stubdb1 now cleans up the temporary file it creates. Sat Sep 14 12:54:52 BST 2002 Olly Betts <olly@survex.com> * docs/Makefile.am: fix "mv apidoc/latex/refman.ps avidoc.ps" which fails after recent change. Sat Sep 14 11:15:33 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Renamed simple* programs in quickstart docs to quickstart* to avoid confusion with the simple* programs in xapian-examples. (Thanks to Alex Bowley for pointing out the confusion.) Sat Sep 14 10:00:35 BST 2002 James Aylett <tartarus@users.sourceforge.net> * configure.in: returned AC_CONDITIONAL for HAVE_PERL Sat Sep 14 09:29:08 BST 2002 Richard Boulton <richard@tartarus.org> * buildall: change to srcdir before testing for necessary tools so that scripts which guess which autotools version to use have more information available to guess with: for example, the Debian autoconf version guessing script looks for "configure.ac" as a sign that autoconf 2.50+ is required. Sat Sep 14 02:50:23 BST 2002 Olly Betts <olly@survex.com> * docs/distributed.txt,docs/quartzdesign.html,docs/todo.xml: assorted documentation updates. Sat Sep 14 02:10:12 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS: removed note about VPATH problems. Sat Sep 14 02:02:58 BST 2002 Olly Betts <olly@survex.com> * btree.cc,btree.h: rearranged Btree::del() so control flow is clearer. Sat Sep 14 01:56:44 BST 2002 Olly Betts <olly@survex.com> * HACKING,docs/Makefile.am: added evil hack to allow VPATH builds to work with BSD make. Wed Sep 11 11:29:07 BST 2002 Olly Betts <olly@survex.com> * backends/database_builder.cc: Quartz is fairly mature - change comment which suggests that the btree names will change. Wed Sep 11 11:13:43 BST 2002 Olly Betts <olly@survex.com> * configure.in: Don't double quote `$MISSING foo' as autoconf adds them * docs/Makefile.am: Use `test a = b' rather than `test a == b' Tue Sep 10 14:50:16 BST 2002 Olly Betts <olly@survex.com> * configure.in: tweaked test for perl modules - stderr is now fully suppressed on sh (which behaves slightly differently to bash). Tue Sep 10 14:27:15 BST 2002 Richard Boulton <richard@tartarus.org> * backend_manager.cc: Fix stub database code so that it doesn't use the gnu extension "getline()". Tue Sep 10 13:51:20 BST 2002 Olly Betts <olly@survex.com> * quartz_table.cc: don't form tag when we just want to check if a key is present before calling Btree::del(). Working out why we can't call Btree::del() would still be worthwhile though. Tue Sep 10 13:02:04 BST 2002 Olly Betts <olly@survex.com> * quartztest.cc: removed unused #include-s; describe valid options in usage message. Tue Sep 10 02:07:29 BST 2002 Olly Betts <olly@survex.com> * remote backend: implemented term_exists() and get_termfreq(); added test to check they work. Mon Sep 9 12:59:15 BST 2002 Richard Boulton <richard@tartarus.org> * Add stub databases: if backend type is auto, and auto_dir points to a file, read in settings from the file (overwriting the existing ones), and then use them to open the database. This allows, for example, omega to open a remote database by putting in a stub database file in the omega directory. Updated documentation for opening databases in userman.html. * apitest: Added test for stub databases. Mon Sep 09 02:26:36 BST 2002 Olly Betts <olly@survex.com> * Documentation updates: move suggested books into intro_ir.html and added URL for citeseer reference Richard posted to mailing list recently; clarified wording in a couple of places. Sat Sep 7 12:52:12 2002 James Aylett <tartarus@users.sourceforge.net> * docs: fixed references to OmDocumentContents (now OmDocument) which had clearly been broken for a while. Also fixed get_data() and set_data() example usage so it reflects reality (I hope). Wed Sep 04 22:46:17 BST 2002 Olly Betts <olly@survex.com> * Corrected min_item.wt to min_wt in two assertions. Thu Aug 15 11:42:20 BST 2002 Richard Boulton <richard@tartarus.org> * buildall: Only add the directory derived from xapian-config to ACLOCAL_FLAGS if it actually exists. Mon Aug 12 01:13:55 BST 2002 Richard Boulton <richard@tartarus.org> * quartz_table.cc: Check that a key exists before calling Btree::del() - there seems to be a bug in the btree code causing a failure when deleting keys if they don't exist. This work around appears to make things work correctly, but won't be terribly efficient. Mon Aug 12 00:09:21 BST 2002 Richard Boulton <richard@tartarus.org> * quartz_table_manager.cc: When unable to apply modifications, call cancel on the buffered table after reopening disk table at old revision - otherwise, the entry count gets messed up. * quartz_table.cc, quartz_database.cc: Improve debugging and error reporting: I'm trying to work out why the failure Olly reported is occurring. Sat Aug 10 15:12:42 2002 James Aylett <tartarus@users.sourceforge.net> * Fixed missing include in quartzcheck.cc Sat Aug 10 14:57:17 2002 James Aylett <tartarus@users.sourceforge.net> * Fixed typo in quartzcheck.cc Tue Jul 23 22:12:34 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Added a note to the effect that the indexgraph stuff isn't really considered useful to the HTML document introducing it. Tue Jul 23 22:09:39 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Added some comments in quartztest test_create1(), in case anyone else who didn't write it needs to understand it. * Added Olly as an active participant to the AUTHORS file :) Tue Jul 23 21:59:50 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Reworded quickstart document so it does not refer to the example code (it was never actually in sync). Mon Jul 22 14:07:18 BST 2002 Sam Liddicott <sam@ananova.com> * Forgot to actually add the quartzcheck.cc file last time Mon Jul 22 13:28:09 BST 2002 Sam Liddicott <sam@ananova.com> * Add buildprereq to package Add quartzcheck which runs btree::check to backends Tue Jul 16 09:44:56 BST 2002 Sam Liddicott <sam@ananova.com> * Fix banding problem; bands are now based on the percentage not the weight, and using integer maths to avoid problems with ceil() rounding up supposed integers which are really slightly larger due to precision errors Thu Jul 11 00:16:22 BST 2002 Olly Betts <olly@survex.com> * configure.in, docs/Makefile.am: configure fixes for Solaris - clear cached result for fdatasync since it may be wrong if we decide we need to link librt; split SOURCEDOC_SRC into two to avoid hitting 4096 char line length in Solaris sed. Tue Jul 09 17:47:31 BST 2002 Olly Betts <olly@survex.com> * PLATFORMS, tests/Makefile.am: improved internaltest.cc automake rules - they should now work in VPATH builds with Solaris make. Tue Jul 09 03:30:41 BST 2002 Olly Betts <olly@survex.com> * configure.in, docs/Makefile.am: Fixed up apidoc and sourcedoc dependencies so that they work correctly in VPATH builds. Also only use portable find arguments to generate SOURCEDOC_SRC. Sat Jul 06 13:53:27 BST 2002 Olly Betts <olly@survex.com> * xapian.spec.in: Updated in preparation for release. Sat Jul 06 13:04:45 BST 2002 Olly Betts <olly@survex.com> * docs/Makefile.am: Run latex twice to get forward references resolved. Fri Jul 05 18:23:08 BST 2002 Olly Betts <olly@survex.com> * matcher/multimatch.cc: better (perhaps even completely correct) fix for sorting into bands. * matcer/localmatch.cc: removed entirely bogus comment from over-eager cut-and-paste. * tests/api_db.cc: added otherwise pointless statement to deldoc3 which seems to fix occasional miscompilation by GCC 2.95. Fri Jul 05 16:37:12 BST 2002 Olly Betts <olly@survex.com> * docs/Makefile.am: "make dist" now fails if dot (part of graphviz) isn't installed. Fri Jul 05 08:54:33 BST 2002 Sam Liddicott <sam@ananova.com> * bindings/: removed some typemap hacks as swig has better typemaps now Fri Jul 05 08:46:06 BST 2002 Sam Liddicott <sam@ananova.com> * matcher/multimatch.cc: Quick hack to stop 100% being in a band all on its own. The real problem is that bands need to include the integer at the upper end of the range, rather than the lower end. Thu Jul 04 16:23:52 BST 2002 Olly Betts <olly@survex.com> * docs/mkdoc.pl: Fixed to work in VPATH builds. Thu Jul 04 12:42:06 BST 2002 Olly Betts <olly@survex.com> * configure.in, docs/Makefile.am: use "missing" to give more helpful errors when we don't find tools we need. Wed Jul 03 12:39:15 BST 2002 Olly Betts <olly@survex.com> * NEWS: incorporated recent changes into provisional version. * docs/: sorted out make rules for running doxygen; don't tell doxygen to build man pages which we aren't using. Tue Jul 02 16:36:54 BST 2002 Olly Betts <olly@survex.com> * docs/: run through ispell; fixed OM_MOP to OmQuery::OP; other corrections. Fri Jun 28 02:59:49 BST 2002 Olly Betts <olly@survex.com> * Backend "auto" now works when creating an OmWritableDatabase which doesn't already exist - it'll pick a backend which supports writing and is compiled in (currently this means quartz). Thu Jun 27 20:31:13 BST 2002 Richard Boulton <richard@tartarus.org> * backends/quartz/btreetest.cc (sequent1): Add regression test for recently fixed bug with sequential addition. * backends/quartz/z_data/ordnum[+-]: Data files for new test. Thu Jun 27 13:13:37 BST 2002 Olly Betts <olly@survex.com> * Documentation fettling: removed docs/Notes and docs/coding_policy.txt moving relevant content into HACKING, common/postlist.h, docs/todo.xml, and include/om/omtypes.h; install docs as PostScript rather than DVI; include text docs in tarball and install them; "make dist" will now fail if a tool needed to build docs is missing; removed last vestiges of Doc++ support; doc/Makefile no longer uses include (include isn't portable); improved doc/tests.txt. Thu Jun 27 12:10:37 BST 2002 Olly Betts <olly@survex.com> * apitest: doesn't directly use anything from utils.h, so don't #include it. Thu Jun 27 12:05:53 BST 2002 Olly Betts <olly@survex.com> * Removed includetest - it was useful early in development, but isn't really now. Wed Jun 26 22:09:57 BST 2002 Olly Betts <olly@survex.com> * apitest: deldoc2 and deldoc3 now pass when the library is compiled with --enable-debug. Wed Jun 26 19:35:56 BST 2002 Olly Betts <olly@survex.com> * Removed code which causes quartztest to fail with an --enable-debug build. Wed Jun 26 13:55:11 BST 2002 Olly Betts <olly@survex.com> * todo.xml: added note to fix 100% being in its own sort band. Wed Jun 26 13:46:02 BST 2002 Richard Boulton <richard@tartarus.org> * btree.cc: Fix splitting of blocks in sequential mode to ensure that split happens no earlier than the midpoint. Wed Jun 26 12:32:47 BST 2002 Richard Boulton <richard@tartarus.org> * btree.cc: Asserts to check that add_item in sequential mode when splitting a block is not trying to add to the wrong block, or to add to a new block which is not empty (or containing more than one item). Tue Jun 25 17:53:19 BST 2002 Olly Betts <olly@survex.com> * Assert that the value of c is sane in GETINT1, etc. Tue Jun 25 16:30:09 BST 2002 Olly Betts <olly@survex.com> * btreetest.cc: use C++ IO rather than C stdio; removed commented out LFSinsertdelete1 test (it was never written and a test suite which creates 2G files is probably a bit too resource hungry). Tue Jun 25 13:56:01 BST 2002 Olly Betts <olly@survex.com> * btree/quartz: More use of const pointers; more use of C++ strings. Wed Jun 19 08:56:00 BST 2002 Sam Liddicott <sam@ananova.com> * Fixed spec file to build for prefix=/usr not just install there Fri Jun 14 16:35:04 BST 2002 Olly Betts <olly@survex.com> * docs/todo.xml: added 4 new tasks. Fri Jun 14 16:33:08 BST 2002 Olly Betts <olly@survex.com> * backends/quartz/quartz_database.cc: Cosmetic changes. Wed Jun 12 16:31:11 BST 2002 Olly Betts <olly@survex.com> * QuartzDbKey, QuartzDbTag: just use a string instead of a string wrapped in a structure. We've stopped inflicting this on external developers (OmData, OmKey/OmValue) so why inflict it on ourselves? Wed Jun 12 12:26:04 BST 2002 Olly Betts <olly@survex.com> * Btree_item: use C++ strings rather than trying to reimplement them. * Fix a few warnings about unused parameters. Wed Jun 12 02:53:16 BST 2002 Olly Betts <olly@survex.com> * Btree_full_compaction, Btree_close: removed - they just wrap methods of Btree. Wed Jun 12 02:28:30 BST 2002 Olly Betts <olly@survex.com> * Btree_item_create and Btree_item_lose subsumed into Btree_item and ~Btree_item. * Pass C++ strings rather than const char * into various Btree open methods. * Removed pointless wrapper functions for Btree open methods. * Btree_quit removed (just use delete directly). * Btree_create removed (just use Btree::create directly). Wed Jun 12 00:48:52 BST 2002 Olly Betts <olly@survex.com> * form_key is now a method of Btree; more passing of strings rather than pointer/length pairs); buffer overrun check added to btreetest. Tue Jun 11 18:04:08 BST 2002 Olly Betts <olly@survex.com> * More C++ like interfaces (e.g. bool returns, strings rather than pointer/length pairs). Tue Jun 11 16:40:08 BST 2002 Richard Boulton <richard@tartarus.org> * quartz/btree.cc: Fix so that it compiles when BTREE_FULL_DEBUG is on, and added a couple of Assert()s. Tue Jun 11 13:09:39 BST 2002 Olly Betts <olly@survex.com> * Removed a number of btree functions which were simply wrappers for calling methods on their first parameter, and most of which weren't even used! * btree_api.txt: updated to better reflect the object-ized API. Tue Jun 11 02:01:53 BST 2002 Olly Betts <olly@survex.com> * Btree::make_index_item(): Added check for buffer overflow. Tue Jun 11 01:49:51 BST 2002 Olly Betts <olly@survex.com> * Still more... Tue Jun 11 01:08:03 BST 2002 Olly Betts <olly@survex.com> * More btree and quartz tidying. Mon Jun 10 20:33:24 BST 2002 Olly Betts <olly@survex.com> * btree.cc,btree.h: cleaned up further. Mon Jun 10 15:54:49 BST 2002 Olly Betts <olly@survex.com> * Further btree and quartz cleanups. Mon Jun 10 14:46:00 BST 2002 Olly Betts <olly@survex.com> * Fixed typo in recent quartzdump change. Mon Jun 10 13:20:20 BST 2002 Olly Betts <olly@survex.com> * Quartz clean-ups - especially QuartzTermList. Mon Jun 10 12:18:40 BST 2002 Olly Betts <olly@survex.com> * Bcursor_create made a method of class Btree as suggested by a FIXME. Mon Jun 10 01:57:53 BST 2002 Olly Betts <olly@survex.com> * More btree fixes and tidying. Fixed bug introduced by typo in previous change which causes tests to fail. Sun Jun 09 17:50:28 BST 2002 Olly Betts <olly@survex.com> * btree_util.h: use inlines rather than macros. * Fix CompileTimeAssert() to avoid potential clashes. Sun Jun 09 12:46:25 BST 2002 Olly Betts <olly@survex.com> * btree.cc,btree.h: fixed bug caused by local variable masking member variable with same name. Tidied up a lot. Wed Jun 5 12:22:57 BST 2002 Sam Liddicott <sam@ananova.com> * Fix OmQuery::op enum handling. Mon May 27 14:28:06 BST 2002 Sam Liddicott <sam@ananova.com> * Add crash-protected next() to OmMSetIterator * Add ->valid() method to iterator which returns FALSE if the iterator has reached the end * Added namespace fixes for latest cvs swig to be released next month Fri May 24 15:28:18 BST 2002 Sam Liddicott <sam@ananova.com> * Took out the string typemaps as swig now has them in stl.i * Added om_percent to omtypes.i - I wonder if we might just read omtypes.h instead, as swig can * Added OmMSetIterator and made changes to OmMSet Fri May 24 11:46:50 BST 2002 Sam Liddicott <sam@ananova.com> * Now support get_matching_terms via swig I'm not sure how to generally map iterators to php but in cases where we know only a small number of items exist to be iterated over it makes sense to return them as a php array Tue May 21 15:28:40 BST 2002 Richard Boulton <richard@tartarus.org> * multi_postlist.cc (get_doclength): Get the document length from the appropriate sub-postlist, rather than from the database. This causes a huge speedup in some cases, since it avoids having to go to the termlist database for every document considered for an mset to get the document length. Added an AssertParanoid to check that the value from the postlist is the same as that from the database. This problem was noticed because "delve -v" was running several orders of magnitude slower than without -v - it is now of comparable speed. Tue May 21 15:27:37 BST 2002 Richard Boulton <richard@tartarus.org> * quartz_values.cc: Small fixes to enable it to compile with debug turned on. Fri May 17 14:52:45 BST 2002 Olly Betts <olly@survex.com> * OmQueryParser::set_database() method added. * Added disabled code to search for a keyword used in the query. * Added ' and * to the list of "phrase making characters". * To make a phrase, a phrase-maker must be followed by an alphanumeric rather than just a non-space. Fri May 17 14:49:54 BST 2002 Olly Betts <olly@survex.com> * INSTALL: GCC 3.1 release version works, so noted this. * HACKING: Added note about warning-free compilation being desirable and suggesting using "./configure CFLAGS=-Werror" to promote this. * NEWS: Minor updates. Thu May 16 16:57:42 BST 2002 Olly Betts <olly@survex.com> * Minor tweaks to get a warning-less compile with GCC 3.1. Thu May 16 13:36:19 BST 2002 Sam Liddicott <sam@ananova.com> * Clean up the spec file Thu May 16 12:20:09 BST 2002 Olly Betts <olly@survex.com> * Documented match_sort_key. Thu May 16 10:42:47 BST 2002 Sam Liddicott <sam@ananova.com> * A few quick tips from Olly, now the tarfile name and version no. stuff comes from autoconf so the spec file won't need tweaking each time the version number changes Wed May 15 15:48:42 BST 2002 Sam Liddicott <sam@ananova.com> * This can build RPM packages. Get the make-dist tarball and do: rpm -ta xapian-core-0.4.1-cvs.tar.gz and it will make source and binary rpms It doesn't build bindings packages yet but this will come when I build them Wed May 15 13:27:21 BST 2002 Richard Boulton <richard@tartarus.org> * buildall: Allow use of automake 1.6.1 and later, and CVS automake (versions 1.6[a-z]) Wed May 15 11:40:58 BST 2002 Sam Liddicott <sam@ananova.com> * SWIG cvs (php4) now writes all output files to the same directory as the main output file, so need for hacks to move these, also no more clobbering our source files when building in the checkout dir Tue May 14 21:54:02 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Updated PLATFORMS file. Tue May 14 16:48:07 BST 2002 Olly Betts <olly@survex.com> * Removed multiple-include protection from config.h - if it's included more than once it's now a bug and we want to know! Tue May 14 15:25:05 BST 2002 Olly Betts <olly@survex.com> * OmQueryParser: parse acronyms as a single term (so "E.T." -> "ET" and "N.A.T.O" -> "NATO"). Tue May 14 15:23:20 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Fixed config.h includes: source definition (ie not header) files #include <config.h> as their first action. Tue May 14 14:47:00 BST 2002 Olly Betts <olly@survex.com> * Updated todo. * extra/parsequery.yy: Fettled indentation. Tue May 14 13:14:51 BST 2002 Olly Betts <olly@survex.com> * Added test case for new "embedded &" rule, and fix problem bug it revealed. Tue May 14 13:10:33 BST 2002 Olly Betts <olly@survex.com> * OmQueryParser: Parse <word>&<word> as a single term (e.g. AT&T, M&S, A&P). Tue May 14 11:50:03 BST 2002 Olly Betts <olly@survex.com> * When including config.h, always specify it as <config.h>, not "config.h", and always do it first. Both are recommended by the autoconf documentation. Mon May 13 17:17:13 BST 2002 James Aylett <tartarus@users.sourceforge.net> * missing fallback rule all-local in docs/Makefile.am was failing the build on Solaris Mon May 13 15:50:31 BST 2002 Sam Liddicott <sam@ananova.com> * Use new %extends notation for swig. * Update for new omValue usage * Add omqueryparser support Mon May 13 15:12:30 BST 2002 James Aylett <tartarus@users.sourceforge.net> * shipped getopt will now build in absence of system getopt Mon May 13 15:09:17 BST 2002 Olly Betts <olly@survex.com> * configure.in: Moved AH_TOP/AH_BOTTOM to more logical place. * docs/todo.xml: Updated. Mon May 13 14:13:27 BST 2002 James Aylett <tartarus@users.sourceforge.net> * added getopt to DIST_SUBDIRS so distributions work Fri May 10 14:54:06 BST 2002 Olly Betts <olly@survex.com> * getopt/: Checked in files I missed last time. Fri May 10 14:35:52 BST 2002 Olly Betts <olly@survex.com> * Make sure GNU getopt is linked to everything which might need it so we build in non-glibc environments. Fri May 10 13:42:59 BST 2002 Olly Betts <olly@survex.com> * Replaced second match_sort_key optimisation with one which is faster and not flawed. Thu May 09 18:27:30 BST 2002 Olly Betts <olly@survex.com> * match_sort_key now runs at a sensible speed (approx. 20x speed up on my test case). Thu May 09 16:58:11 BST 2002 Olly Betts <olly@survex.com> * Added another optimisation for match_sort_key - my test case is now 25% faster (still slow though). Thu May 09 11:57:59 BST 2002 Olly Betts <olly@survex.com> * NEWS, PLATFORMS, README: Updated to something approximating what we want for a release. * docs/todo.xml: Updated. Thu May 09 10:57:59 BST 2002 Olly Betts <olly@survex.com> * Removed references to match_max_or_terms (which was replaced by OmQuery::OP_ELITE_SET long, long ago). Wed May 08 17:38:50 BST 2002 Olly Betts <olly@survex.com> * Improved api documentation comments. Wed May 08 17:09:14 BST 2002 Olly Betts <olly@survex.com> * Include findheaders.pl in tarball. Wed May 08 16:09:14 BST 2002 Olly Betts <olly@survex.com> * Killed off acconfig.h. Wed May 08 15:53:11 BST 2002 Olly Betts <olly@survex.com> * Added autom4te-*.cache to .cvsignore. Wed May 08 15:50:56 BST 2002 Olly Betts <olly@survex.com> * Rebuild docs on "make", not just "make dist" or "make install". Wed May 08 15:07:01 BST 2002 Olly Betts <olly@survex.com> * Added workaround for problems with <fcntl.h> on Solaris when LFS is enabled. Wed May 08 13:44:10 BST 2002 Olly Betts <olly@survex.com> * Matcher can now sort on a key. Can be rather slow at present... Wed May 08 13:15:21 BST 2002 Olly Betts <olly@survex.com> * Updated todo. Wed May 08 12:22:53 BST 2002 Olly Betts <olly@survex.com> * Removed ltmain.sh-s from CVS. Wed May 08 12:21:32 BST 2002 Olly Betts <olly@survex.com> * buildall: run libtoolize; stripped out unnecessary stuff. Wed May 08 09:44:45 BST 2002 Olly Betts <olly@survex.com> * matcher/msetpostlist.cc: Removed bogus comment cut and pasted from matcher/mergepostlist.cc. Tue May 07 17:51:57 BST 2002 Olly Betts <olly@survex.com> * It's "m4_include", not "m4include". Tue May 07 17:52:36 BST 2002 Olly Betts <olly@survex.com> * Added autoconf/type_socklen_t macro. Tue May 07 16:50:20 BST 2002 Olly Betts <olly@survex.com> * Now require autoconf 2.50, which allows us to clean up some of the more unpleasant parts of the build system. Tue May 07 14:55:37 BST 2002 Olly Betts <olly@survex.com> * Added "match_sort_bands" option to sort results within relevance bands. Currently the sort within each band is by document id, which isn't really very useful. The plan is to allow sorting by a key. Tue May 07 14:54:12 BST 2002 Olly Betts <olly@survex.com> * Sorted out tests/findheaders.pl wrt last change. Tue May 07 13:12:23 BST 2002 Olly Betts <olly@survex.com> * Run perl scripts from Makefiles explicitly with $(PERL) rather than having them as generated files in AC_OUTPUT. Mon May 6 16:04:37 2002 James Aylett <tartarus@users.sourceforge.net> * Distribution fixes: make clean should now clear up all required files. Mon May 6 14:42:46 2002 James Aylett <tartarus@users.sourceforge.net> * Don't include dlfcn.h since we don't use it (and doesn't exist on some target platforms). Mon May 6 14:07:24 2002 James Aylett <tartarus@users.sourceforge.net> * Minor documentation corrections. Fri May 03 18:14:14 BST 2002 Olly Betts <olly@survex.com> * Changed autoconf/automake project name to xapian-core, and updated the documentation to reflect this. * Disabled the bindings from being included in distribution tarballs. Fri May 03 18:13:43 BST 2002 Olly Betts <olly@survex.com> * Updated todo list. Fri May 3 18:02:06 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Overloading ambiguity resolutions for linux/alpha build on CF Fri May 03 17:13:36 BST 2002 Olly Betts <olly@survex.com> * buildall: Don't run configure or make; removed --no-make and --quiet. Fri May 3 16:52:01 BST 2002 James Aylett <tartarus@users.sourceforge.net> * tests/testdata/etext.txt needs to be included in the distribution Fri May 03 16:18:38 BST 2002 Olly Betts <olly@survex.com> * Removed OmValue in favour of simply using a string (analogous to recent OmData change). Fri May 03 14:52:04 BST 2002 Olly Betts <olly@survex.com> * Removed unused om_docname typedef. Fri May 03 14:42:31 BST 2002 Olly Betts <olly@survex.com> * Terminology change - a "key" (in the OmKey sense) is now a "value". Fri May 3 14:31:57 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Fixed up todo list and HTML generation script to cope with items without an assigned priority. Fri May 03 14:29:58 BST 2002 Olly Betts <olly@survex.com> * It's HACKING (not INSTALL) that has details of developer tools. Thu May 02 15:57:48 BST 2002 Olly Betts <olly@survex.com> * Removed OmData - just return and take a string instead. Thu May 02 14:40:21 BST 2002 Olly Betts <olly@survex.com> * docs/.cvsignore: updated. Thu May 02 14:38:36 BST 2002 Olly Betts <olly@survex.com> * Removed assorted references to pthread locking; updated HACKING. Thu May 02 14:15:47 BST 2002 Olly Betts <olly@survex.com> * Stripped out OmBatchEnquire stuff, as it's unlikely to get sorted out any time soon, and it can be extracted from CVS when someone wants it. Thu May 02 13:50:04 BST 2002 Olly Betts <olly@survex.com> * Check for bison >= 1.35 in buildall. Wed May 01 17:39:57 BST 2002 Olly Betts <olly@survex.com> * Removed lingering references to Omsee and OpenMuscat. Wed May 01 16:34:13 BST 2002 Olly Betts <olly@survex.com> * Removed unused methods from NetClient (read_data and write_data). Wed May 01 14:01:12 BST 2002 Olly Betts <olly@survex.com> * Added test consistency1 to check matcher optimisations don't result in inconsistent results (i.e. that same matches and weights are returned regardless of size of mset asked for). Wed May 01 11:48:10 BST 2002 Olly Betts <olly@survex.com> * Fixed mismatched tags in todo.xml. Wed May 01 11:46:42 BST 2002 Olly Betts <olly@survex.com> * Updated todo. Wed May 01 11:40:47 BST 2002 Olly Betts <olly@survex.com> * Corrected configure warning message - it's XML::Parser not XML. Wed May 01 11:34:49 BST 2002 Olly Betts <olly@survex.com> * Tweaks to get queryparser building cleanly. Tue Apr 30 17:44:48 BST 2002 Olly Betts <olly@survex.com> * Updated query parser in extra from the updated one in omega, and fettled it to cope better with multiple instantiations (though it's a long way from being reentrant at the moment). Tue Apr 30 11:03:44 BST 2002 Olly Betts <olly@survex.com> * Updated todo. Mon Apr 29 19:28:04 BST 2002 Olly Betts <olly@survex.com> * More string.c_str() tidying. Mon Apr 29 17:41:33 BST 2002 Olly Betts <olly@survex.com> * Tidied up some uses of string.c_str(). * Cleaned up todo list. Mon Apr 29 16:08:21 BST 2002 Olly Betts <olly@survex.com> * Updated buildall to check various tools have the required version numbers. Mon Apr 29 15:58:57 BST 2002 Olly Betts <olly@survex.com> * Tweaked use of HAVE_STREAMBUF Fri Apr 26 14:41:09 BST 2002 Sam Liddicott <sam@ananova.com> * Swig is now modified so it can build config.m4 and Makefile.in for php modules so I removed these from bindings/php4 Tue Apr 23 18:21:03 BST 2002 Olly Betts <olly@survex.com> * Tried compiling with GCC 3.1 prerelease snapshot and fixed various STL-related warnings (mostly uses of istrstream). * Fixed usage of AM_CFLAGS and AM_CXXFLAGS. Tue Apr 23 15:48:36 BST 2002 Olly Betts <olly@survex.com> * Updated buildall (and had to tweak it again, sigh). Tue Apr 23 14:21:01 BST 2002 Sam Liddicott <sam@ananova.com> * Added typemap conversion to generate OmSettings from php hash * Changed order of classes in interface file Tue Apr 23 14:20:05 BST 2002 Olly Betts <olly@survex.com> * Revamped xapian-config and xapian.m4. Important change - it's now "xapian-config --cxxflags" and XAPIAN_CXXFLAGS, not --cflags and _CFLAGS. Mon Apr 22 17:58:56 BST 2002 Olly Betts <olly@survex.com> * Removed leak checker. Mon Apr 22 13:40:01 BST 2002 Olly Betts <olly@survex.com> * Tiny code tweak. Mon Apr 22 13:11:32 BST 2002 Olly Betts <olly@survex.com> * Write debug log using unbuffered IO and O_APPEND which should prevent the log file becoming garbled when used from applications which use Xapian objects in multiple threads. Mon Apr 22 12:19:18 BST 2002 Olly Betts <olly@survex.com> * Updated copyright notices for last change. Mon Apr 22 11:20:28 BST 2002 Olly Betts <olly@survex.com> * Removed thread locks. Fri Apr 19 17:38:23 BST 2002 Olly Betts <olly@survex.com> * Removed completed tasks from todo list. Fri Apr 19 17:26:23 BST 2002 Olly Betts <olly@survex.com> * Removed placeholder code for OmQuery::OP_PERCENT_CUTOFF - it's actually impossible to implement since we started rescaling percentages so that 100% is attainable. Fri Apr 19 16:42:42 BST 2002 Olly Betts <olly@survex.com> * Tracked down and fixed the bug in the quartz backend which was making test poslist2 fail. Fri Apr 19 16:06:39 BST 2002 Olly Betts <olly@survex.com> * Fixed some compilation problems which I somehow failed to notice before the last check-in. Fri Apr 19 15:04:13 BST 2002 Olly Betts <olly@survex.com> * Added more RETURN(...) wrappers. Fri Apr 19 14:46:19 BST 2002 Olly Betts <olly@survex.com> * Added more DEBUGCALL tracing to quartz. Fri Apr 19 13:29:57 BST 2002 Olly Betts <olly@survex.com> * Minor tweaks to quartz_database.cc. Fri Apr 19 13:07:50 BST 2002 Olly Betts <olly@survex.com> * Moved test poslist1 into correct place and renamed it to poslist2 since there's already a poslist1... Fri Apr 19 11:19:15 BST 2002 Olly Betts <olly@survex.com> * poslist1 (when actually run!) reveals that Quartz fails to throw errors in this case. Disabled test for now, as I can't see why it fails to throw... Thu Apr 18 17:44:56 BST 2002 Olly Betts <olly@survex.com> * Added poslist1 test as suggested by a TODO entry. Thu Apr 18 14:17:24 BST 2002 Olly Betts <olly@survex.com> * buildall: Removed search for GNU make (as of automake 1.5 it's not needed for automake and we aim to have portable makefiles). Made buildall more portable. Thu Apr 18 14:16:18 BST 2002 Olly Betts <olly@survex.com> * xapian-config.nodep is a generated file, so don't look for it in srcdir. Thu Apr 18 12:45:04 BST 2002 Olly Betts <olly@survex.com> * Added bindings/php4/xapian/.cvsignore. Thu Apr 18 12:36:53 BST 2002 Olly Betts <olly@survex.com> * Minor tweaks to "Makefile" section in HACKING. Thu Apr 18 12:26:21 BST 2002 Olly Betts <olly@survex.com> * Yet more Makefile portability fixes; added section to HACKING summarising my recent experiences. Thu Apr 18 11:26:37 BST 2002 Olly Betts <olly@survex.com> * Corrected the sense of the test in the previous check-in. Thu Apr 18 11:08:04 BST 2002 Olly Betts <olly@survex.com> * Use case instead of echo/sed to replace dirname. Thu Apr 18 10:19:01 BST 2002 Olly Betts <olly@survex.com> * Another Makefile portability tweak. Wed Apr 17 16:07:02 BST 2002 Olly Betts <olly@survex.com> * bindings/php4/Makefile.am: Fixed a couple of problems with recent changes. Wed Apr 17 15:32:44 BST 2002 Olly Betts <olly@survex.com> * Makefile portability improvements. Wed Apr 17 14:29:21 BST 2002 Olly Betts <olly@survex.com> * Replaced uses of GNU make specific features with more portable ones. Wed Apr 17 12:43:07 BST 2002 Olly Betts <olly@survex.com> * Update todo list. Wed Apr 17 12:41:03 BST 2002 Olly Betts <olly@survex.com> * Removed BrightStation's logo from the documentation. Wed Apr 17 12:27:07 BST 2002 Olly Betts <olly@survex.com> * xapian-config: Removed --uninst support - it's very hard to make it work reliably and we're better off directing our efforts towards improving the library than trying to get it working and keep it working. Wed Apr 17 10:39:02 BST 2002 Olly Betts <olly@survex.com> * Fixed btreetest code to use new names "ord+" and "ord-" for datafiles; disabled LFSinsertdelete1 test as it's exactly the same as insertdelete1 (despite the comment above which suggests it tests files >2G). Mon Apr 15 17:14:22 BST 2002 Olly Betts <olly@survex.com> * Fix a few OmDatabase::InternalInterface uses I'd somehow missed. Mon Apr 15 16:08:20 BST 2002 Olly Betts <olly@survex.com> * Mark internal classes as @internal for doxygen; removed all uses of OmDatabase::InternalInterface as it's no longer needed. Mon Apr 15 15:42:47 BST 2002 Olly Betts <olly@survex.com> * Updated TODO list. Mon Apr 15 2002 Sam Liddicott <sam@ananova.com> * Hacked around bindings and bindings/php to remove some SWIG workarounds. SWIG cvs + some new non-committed patches is need still. Mon Apr 15 12:12:05 BST 2002 Olly Betts <olly@survex.com> * stemtest: getopt fix. * stemtest.pl.in: output now: less verbose by default and coloured. Mon Apr 15 11:18:58 BST 2002 Olly Betts <olly@survex.com> * Updated HACKING to say that automake 1.5 is required, and note that automake 1.6.1 is better than 1.6. Sat Apr 13 20:35:51 2002 James Aylett <tartarus@users.sourceforge.net> * Fixed stemtest getopt usage so it works again. Sat Apr 13 20:20:52 BST 2002 James Aylett <tartarus@users.sourceforge.net> * Fixed omtcpsrv getopt usage so it works again. Fri Apr 12 16:24:06 BST 2002 Olly Betts <olly@survex.com> * Require automake 1.5. Fri Apr 12 12:11:57 BST 2002 Olly Betts <olly@survex.com> * omtcpsrv: Namespace fixes. Fri Apr 12 12:25:42 BST 2002 Olly Betts <olly@survex.com> * Fixed everything to use getopt rather than icky handparsing of options. Fri Apr 12 11:26:57 BST 2002 Olly Betts <olly@survex.com> * Removed delve - it's going to be an example program. Thu Apr 11 18:09:35 BST 2002 Olly Betts <olly@survex.com> * Command line argument improvements - added FIXMEs where getopt should be used. Thu Apr 11 17:01:44 BST 2002 Richard Boulton <richard@tartarus.org> * Fixes to work with automake 1.6.1. I recommend using 1.6.1 rather than 1.6, since this fixes many bugs in 1.6. I havn't tested with 1.6 recently, but it is likely to report problems. Thu Apr 11 17:25:37 BST 2002 Olly Betts <olly@survex.com> * Use GNU getopt. Thu Apr 11 15:06:15 BST 2002 Olly Betts <olly@survex.com> * Updated to recommend automake 1.5, noting problem with automake 1.4 and that we'll probably soon insist on 1.5, and later autoconf 2.50. Also added notes on recommended and usable GCC versions. Thu Apr 11 14:23:44 BST 2002 Olly Betts <olly@survex.com> * Cleaning up build system. Thu Apr 11 13:18:23 BST 2002 Olly Betts <olly@survex.com> * Use AndPostList with boolean weights on the RHS instead of FilterPostList - this appears to fix a bug, but I suspect it's actually just moving it around. Will investigate later. Thu Apr 11 10:11:11 BST 2002 Olly Betts <olly@survex.com> * Sorted out automake for php4 bindings. Wed Apr 10 14:44:50 BST 2002 Olly Betts <olly@survex.com> * Removed a superfluous DEBUGLINE. Tue Apr 9 19:02:09 BST 2002 Richard Boulton <richard@tartarus.org> * Fix some DEBUGLINE() macros in the matcher which had the first parameter missing. Tue Apr 09 17:50:44 BST 2002 Olly Betts <olly@survex.com> * Removed stray bit of debug code. Tue Apr 09 17:05:09 BST 2002 Olly Betts <olly@survex.com> * More debug tracing added. Tue Apr 09 10:48:42 BST 2002 Olly Betts <olly@survex.com> * Disabled pthread support by default. Sat Apr 6 20:45:21 2002 James Aylett <tartarus@users.sourceforge.net> * Reordered some output generation in configure so it won't complain about directories not existing. (I don't understand what the problem is, but this does fix it.) Sat Apr 06 20:27:57 BST 2002 Olly Betts <olly@survex.com> * Moved a few lingering traces of XML indexer stuff to indexer subdirectory. Sat Apr 06 17:37:02 BST 2002 Olly Betts <olly@survex.com> * Removed -pedantic (read the gcc docs - it's doesn't turn on useful warnings) and fixed all the remaining compilation warnings in a non-debug build. Fri Apr 05 10:21:28 BST 2002 Olly Betts <olly@survex.com> * Removed --enable-profiling, --enable-purify, and --enable-insure. They don't do anything which can't be achieved by passing environment variables and switches to configure (now documented in HACKING). Adding a separate switch for every profiling and code-quality tool out there isn't a sensible approach. Thu Apr 04 20:37:12 BST 2002 Olly Betts <olly@survex.com> * Updated .cvsignore files. Thu Apr 04 20:25:41 BST 2002 Olly Betts <olly@survex.com> * PTHREAD_CFLAGS was only being used for compiling C, not C++, which was causing builds to fail unless threading was disabled. This problem has been there a long time, but until recently was hidden by the check for fdatasync erroneously linking in -lrt which pulls in pthreads anyway (at least this is the situation on Linux). Thu Apr 04 20:06:00 BST 2002 Olly Betts <olly@survex.com> * Removed XML indexer stuff from the main xapian library. All the removed bits are now in the indexer subdirectory, ready to be reassembled into a supplementary library. Thu Apr 04 15:43:43 BST 2002 Olly Betts <olly@survex.com> * More debug tracing in the matcher. Thu Apr 04 14:17:10 BST 2002 Olly Betts <olly@survex.com> * Added debug tracing to most of the matcher. Thu Mar 28 17:21:16 GMT 2002 Olly Betts <olly@survex.com> * Give up trying to make internals private or protected in the header files in include - the consequences are just too ugly, and it's clear enough that they shouldn't be messed with except by the library itself. Wed Mar 27 10:24:26 GMT 2002 Olly Betts <olly@survex.com> * Don't add "-lrt" to the link line unless it's actually needed for fdatasync(). Wed Mar 27 10:11:09 GMT 2002 Olly Betts <olly@survex.com> * Fixed two problems in the matcher which were respectively causing problems with boolean filters, and with situations where OR or ANDMAYBE decayed to AND. Fri Jan 11 18:00:44 GMT 2002 James Aylett <tartarus@users.sourceforge.net> * Various build fixes for libxml2: define CHAR as needed, --enable-validation (off by default), check for and discard 'blank' text/cdata nodes as we encounter them Fri Jan 11 15:40:50 GMT 2002 Sam Liddicott <sam@ananova.com> * Added Large File Support in the same way it was added to GNU tar. (If it's good enough for them...) This is by means of autoconf/ac_sys_largefile which I'm told is built in to later versions of autoconf, and may possibly give us trouble redefining it for those versions... But we'll see about that when/if it happens. It seems to work fine here on some DB's which I had to stop growing cos they got too big; I'm now past the 2G boundary with no problems I guess we should still add overflow detection for 2^31 blocks. Ideas? I need some help on the btreetest.cc as I haven't yet worked out what it does; so while I've added a test, it doesn't yet test what it says it does. I have changed btreetest.cc so you can tell it where to dump the test DB Fri Dec 21 17:48:09 GMT 2001 Olly Betts <olly@survex.com> * Fixed halflife calculation in OmBiasFunctor. Fri Dec 21 15:24:58 GMT 2001 Olly Betts <olly@survex.com> * Added a temporary API to allow use of OmBiasFunctor. Fri Dec 21 12:58:23 GMT 2001 Olly Betts <olly@survex.com> * Disable conversion of OR to AND/ANDMAYBE in matcher tree - there's a bug which sometimes manifests when there's a date filter and a threshold cutoff in omega, and results in just one match when there should be more. I don't believe the bug is actually in OrPostList but I've been tracking it for 4 days without success and turning off this optimisation seems to stop it occurring so will have to do for now. Thu Dec 20 17:41:04 GMT 2001 Olly Betts <olly@survex.com> * matcher/multimatch.cc: added more percent_cutoff assertions. Thu Dec 20 13:41:41 GMT 2001 Olly Betts <olly@survex.com> * Fixed incorrectly classified debug message; fixed compiler warning; more comments about matches_* with a percent_cutoff. Thu Dec 20 13:09:51 GMT 2001 Olly Betts <olly@survex.com> * Better OmPostListIterator::get_description() (call get_description on internal->postlist). Thu Dec 20 12:42:55 GMT 2001 Olly Betts <olly@survex.com> * Fixed assorted problems with omtcpsrv and co from recent change. Thu Dec 20 10:41:57 GMT 2001 Richard Boulton <richard@tartarus.org> * In ./configure --help, correctly report muscat36 backend as defaulting to off. Thu Dec 20 10:36:18 GMT 2001 Olly Betts <olly@survex.com> * Don't delete internal in ~OmWritableDatabase - ~OmDatabase does that for us (internal was zeroed after delete, so this was probably harmless) * Tweaked OmEnquire::get_description() and OmPostListIterator::get_description() to do less - there's a debug related bug somewhere and get_description() should be a non-intrusive method... Thu Dec 20 10:32:56 GMT 2001 Richard Boulton <richard@tartarus.org> * Fix a couple of misquoted messages in configure.in. Tue Dec 18 23:22:00 2001 James Aylett <tartarus@users.sourceforge.net> * Use xapian.org where appropriate in documentation references Tue Dec 18 14:32:32 2001 James Aylett <tartarus@users.sourceforge.net> * Use xapian.org not sourceforge for schemas Mon Dec 17 18:47:08 GMT 2001 Olly Betts <olly@survex.com> * Cured problems with string('x') caused by overzealous search and replace on my part. Mon Dec 17 16:44:31 GMT 2001 Olly Betts <olly@survex.com> * Improved matches_estimated when a percent_cutoff is in effect. Sun Dec 16 18:09:28 GMT 2001 Olly Betts <olly@survex.com> * You can now register a match decider functor by name with an OmEnquire object. It's not yet used for anything though. Sun Dec 16 17:31:08 GMT 2001 Olly Betts <olly@survex.com> * omtcpsrv: simplified arguments - just list database directories and they'll be opened with the auto backend. Removed --im as it didn't work as advertised and nobody's noticed (as actually implemented it was useless). Sat Dec 15 14:10:32 GMT 2001 Olly Betts <olly@survex.com> * Minor update to todo.xml. Sat Dec 15 13:43:47 GMT 2001 Olly Betts <olly@survex.com> * More fettling of OmBiasFunctor stuff - still not ready for use yet... Tue Dec 11 13:06:08 GMT 2001 Olly Betts <olly@survex.com> * Don't work out probabilistic weights for boolean bits of the query (rhs of FILTER and AND_NOT) which fixes incorrect percentage weights in filtered queries. Tue Dec 11 12:26:44 GMT 2001 Olly Betts <olly@survex.com> * Added test singlesubq1 to keep an eye on can_replace_by_single_subq. Mon Dec 10 16:58:21 GMT 2001 Olly Betts <olly@survex.com> * can_replace_by_single_subq shouldn't return true for OP_ELITE_SET since you can't call set_elite_set_size() on an arbitrary subquery. Thu Dec 6 15:02:05 GMT 2001 Richard Boulton <richard@tartarus.org> * (omqueryinternal.cc, omtermlistiteratorinternal.h) Fix some compiler warnings (when compiling with no debug) due to methods which should return a value but simply contain "Assert(false);". Throw an exception in such methods instead (after the Assert). Mon Dec 03 11:20:42 GMT 2001 Olly Betts <olly@survex.com> * New BiasPostList - not yet enabled as it needs more work. Mon Dec 03 11:16:44 GMT 2001 Olly Betts <olly@survex.com> * quartzdump: give syntax error if no tables specified. Thu Nov 29 14:10:43 GMT 2001 Olly Betts <olly@survex.com> * ExtraWeightPostlist wasn't handling prune correctly. Thu Nov 22 13:46:59 GMT 2001 Olly Betts <olly@survex.com> * Fixed segfault when percentage_cutoff emptied the proto-mset. Thu Nov 22 12:27:49 GMT 2001 Olly Betts <olly@survex.com> * Enhanced test emptyquery1. Fri Nov 16 15:32:09 GMT 2001 Olly Betts <olly@survex.com> * Added OmQuery::is_empty() method. Thu Nov 15 15:36:07 GMT 2001 Olly Betts <olly@survex.com> * Updated todo list. Thu Nov 15 13:03:05 GMT 2001 Olly Betts <olly@survex.com> * Fixed percent_cutoff problems. Wed Nov 14 12:05:44 GMT 2001 Olly Betts <olly@survex.com> * Empty OmQuery() now matches no documents (rather than causing an exception). Thu Nov 08 18:23:21 GMT 2001 Olly Betts <olly@survex.com> * SIGSTKFLT isn't portable, so added #ifdef checks for it. Wed Nov 07 15:34:29 GMT 2001 Olly Betts <olly@survex.com> * Fixed 2 GCC 3.0 warnings in muscat36 backend. Wed Nov 07 14:21:47 GMT 2001 Olly Betts <olly@survex.com> * Made doc counts returned from percentage cutoff matches more accurate. Wed Nov 07 12:30:32 GMT 2001 Olly Betts <olly@survex.com> * GCC 3.0 fixes. Tue Nov 06 11:16:42 GMT 2001 Olly Betts <olly@survex.com> * Matcher now calculates percentages like Omega does (or did as I'm about to remove that code from Omega), rather than as a percentage of the theoretical maximum possible weight. The old way generally gives disappointingly low values, and by pushing the percentage code down into the matcher, it can be used to optimise the query. Tue Nov 06 11:13:06 GMT 2001 Olly Betts <olly@survex.com> * Quick fix for problems with null OmQuery-s. Mon Nov 05 15:52:55 GMT 2001 Olly Betts <olly@survex.com> * Added signal handling to testsuite - if a testcase generates a signal it is caught and reported, then the testsuite continues with further testcases. Wed Oct 31 14:26:59 GMT 2001 Olly Betts <olly@survex.com> * Removed needless complications in matcher when a weight threshold has been set. Tue Oct 30 16:50:46 GMT 2001 Olly Betts <olly@survex.com> * Removed superfluous lines from configure.in; fixed warning from automake 1.5. Mon Oct 29 11:54:36 GMT 2001 Olly Betts <olly@survex.com> * Changed matcher to form proto-mset using a min-heap (as described in "Managing Gigabytes", 2nd ed., pp 211-213). Compared to the previous "nth-element" approach, this uses about half the memory, allows the min-weight based matcher optimisations to work sooner, and should be inherently faster (not yet benchmarked though). The code's slightly simpler too. Mon Oct 29 11:53:38 GMT 2001 Olly Betts <olly@survex.com> * Fixed remote backend to pass match_cutoff setting across. Fri Oct 26 13:43:08 BST 2001 Olly Betts <olly@survex.com> * Reserve size needed for mset vector. Thu Oct 25 18:00:54 BST 2001 Olly Betts <olly@survex.com> * Fixed a couple of stray references to the old sleepycat backend. Tue Oct 23 18:19:32 BST 2001 Olly Betts <olly@survex.com> * More cleaning up of MultiMatch. Tue Oct 23 16:14:27 BST 2001 Olly Betts <olly@survex.com> * Removed RemotePostList - it was an experimental idea which proved to be too slow even on a fast network. Cleaned up code where hooks were bodged in for it. Tue Oct 23 13:27:12 2001 James Aylett <tartarus@users.sourceforge.net> * Added documentation of the test system (albeit brief). Also mentioned the autotools book in HACKING. Tue Oct 23 12:07:59 BST 2001 Olly Betts <olly@survex.com> * Updated .cvsignore files. Tue Oct 23 11:29:47 BST 2001 Olly Betts <olly@survex.com> * Code tidying while investigating sporadic fails from quartztest. Mon Oct 22 17:42:30 BST 2001 Olly Betts <olly@survex.com> * Minor fettles to quartztest: fixed warnings, whitespace tidying, use mkdir(...) rather than system("mkdir ...") Mon Oct 22 16:41:12 2001 James Aylett <tartarus@users.sourceforge.net> * Renamed project to Xapian. Mon Oct 22 14:22:00 BST 2001 Olly Betts <olly@survex.com> * Fixed non-ISO C++ code so Xapian will compile under GCC 3.0 (and also RedHat's "GCC 2.96"). Mon Oct 22 11:21:45 2001 James Aylett <tartarus@users.sourceforge.net> * Altered omseek-config and the AC m4 macro to make uninst executables link to the right version of the library Mon Oct 22 10:38:20 2001 James Aylett <tartarus@users.sourceforge.net> * Minor fixes to enable build (and test) under Solaris. Fri Jun 22 13:24:59 2001 Sam Liddicott <sam@ananova.com> * Added some explicit type casts in xmlindexer and quartz backend to help gcc2.96 compile. Thu Jun 21 11:05:31 2001 James Aylett <tartarus@users.sourceforge.net> * HACKING updated to (a) remove inaccurate statements about copyright assignment, and (b) talk about developer write access to CVS. Mon Jun 11 13:24:43 2001 James Aylett <tartarus@users.sourceforge.net> * pthread support detected using an automake macro. Should be more portable as a result. * We need GNU make: buildall now tries to find it. Mon Jun 11 12:07:00 2001 Hein Ragas <hragas@users.sourceforge.net> * Modified QuartzWritableDatabase::do_add_document and QuartzWritableDatabase::do_replace_document to leave the PositionList empty for a term with no positions associated with it. Mon May 29 12:17:00 2001 Hein Ragas <hragas@users.sourceforge.net> * Modified QuartzPostList::add_entry to allow for adding of entries in the middle of the postlist, not only at the end. Mon May 28 12:00:00 2001 Hein Ragas <hragas@users.sourceforge.net> * Fixed a bug in replace_document for Quartz, added a test to ensure proper working of the fix. Fri Apr 27 15:16:10 2001 James Aylett <tartarus@users.sourceforge.net> * Project name change to Omseek. Wed Apr 11 09:06:37 BST 2001 Chris Emerson <chris.emerson@omsee.com> * Implemented deleting of items from postlists, and added some tests to check that it works properly. * Fixed a potential memory leak in QuartzPostList if an exception was thrown from the constructor. * BackendManager now sets the quartz_logfile setting, to make the logs available from the testsuite. * Added mention of quartz_logfile to docs/omsettings * Removed a bad assert from InMemoryDatabase::get_doclength() Mon Apr 9 21:48:18 BST 2001 Chris Emerson <chris.emerson@omsee.com> * Added #include <cmath> to testutils.cc Fri Mar 30 17:37:24 BST 2001 Olly Betts * Added extra/.cvsignore Fri Mar 30 17:22:13 BST 2001 Olly Betts * Disabled allocation checking of new[] and delete[] since some STL implementations use them which leads to leaks being reported where none exist. This is addressable longer term... Thu Mar 29 19:05:46 BST 2001 Richard Boulton <richard.boulton@omsee.com> * Add "extra" directory to om source tree, to hold extra libraries for use with omsee, which aren't part of the core. * Put a query parser into "extra". This is roughly the parser from omega. Needs some work: in particular, it needs to be made reentrant. * Added extra/ directory to top level Makefile.am and configure.in. Added AC_PROG_YACC to Makefile.am * Add includes of required header files to omquery.h. Wed Mar 28 17:15:22 BST 2001 Richard Boulton <richard.boulton@omsee.com> * Modifications to OmQuery. Remove the isbool flag: there is no longer any need for the concept of a "pure boolean query". The same effect as was obtained by setting isbool can be obtained by setting the match_weighting_scheme parameter to bool. * Also, undefined queries may no longer be used when building up composite queries; the correct behaviour for this is hard to define and implement and harder to document, and removing this removes lots of special case code. * Updated network query passing code. * Remove special cases for bool weighting in localmatch and irweight. * Updated omsettings docs. Updated comments in omsetting.h. * Updated todo.xml * Incremented network protocol version. * Added test_emptyop1, which checks what happens when an attempt is made to use an empty list of subqueries to build up a query, and then use the query. * Remove test_boolsubq1: no such thing as a boolean query now, so no need to test what happens when one is used. * Updated testsuite to use "match_weighting_scheme" setting, rather than calling OmQuery::set_bool(). Wed Mar 28 15:05:46 BST 2001 Chris Emerson <chris.emerson@omsee.com> * Fixes to InMemoryDatabase: + do_open_post_list returns an EmptyPostList rather than an assertion error if the term doesn't exist + Fixed doccount(), which wasn't decrementing on delete_doc() + Remove terms from the database when they have no more postings. + Implement collection frequencies, and run collfreq tests for inmemory. Wed Mar 28 14:24:38 BST 2001 Richard Boulton <richard.boulton@omsee.com> * Set elite set size to max(10, sqrt(number of subqueries)) by default. This is rather more useful than the old default of 0. Wed Mar 28 10:25:35 BST 2001 Chris Emerson <chris.emerson@omsee.com> * QuartzDatabase now implements replace_document(), and delete_document() now removes keys/attributes as well. However, postlist entries are not yet correctly removed, so this will cause problems. * Added a test for deleting/replacing documents, which passes. Tue Mar 27 17:31:58 BST 2001 Chris Emerson <chris.emerson@omsee.com> * QuartzAttributes now has a method to delete a document's attributes, and QuartzRecordManager has a replace_ method. (But the code which uses these isn't in this commit). Thu Mar 22 15:58:08 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Change definition of OmQuery::is_defined(), so that elite set queries are undefined until a size is specified for the set. Wed Mar 21 14:29:04 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Implemented delete_document and replace_document in inmemory database. Wed Mar 21 12:39:13 GMT 2001 Olly Betts * Tweak to docs/quickstart.html. Wed Mar 21 02:50:19 GMT 2001 Olly Betts * Entirely removed sleepcat backend and all references to it. * Disable muscat36 backend by default. Tue Mar 20 18:47:12 GMT 2001 Olly Betts * open_document() now takes a lazy flag - if not set we need to check that the docid given actually exists. * heavy_duty flag now a bool instead of an int. Mon Mar 19 19:16:34 GMT 2001 Olly Betts * Added test getdoc1 to check exceptions are thrown for out-of-range docids (currently they aren't for quartz and muscat36 backends). Mon Mar 19 17:17:31 GMT 2001 Olly Betts * New simpler implementation of leak detection in testsuite. Currently doesn't support malloc/calloc/realloc/free. Mon Mar 19 16:46:16 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Further small modification to omsee-config: will now look in .libs subdirectories, too. Mon Mar 19 15:57:18 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Update omsee-config to work when a prefix is specified that points directly to the directory with the library, rather than expecting the library to be in a subdir "libs/". (libs/ is still used if it exists). * Update omsee.m4, to use omsee-config even when using an uninstalled version of omsee. This in turn fixes problems using uninstalled versions of omsee with STLport. Fri Mar 16 18:14:52 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Updated licenses, to include year 2001. Fri Mar 16 15:40:49 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Fix up small bug in quartztest causing it to fail for lack of an expected exception. Initialise values in quartz_positionlist.cc when position list is not present. Thu Mar 15 17:40:38 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Make timeout absolute, rather than cumulative, when fetching multiple documents from database: if two nodes fail, the total timeout should still be that specified, not twice it. * Make the timeout set by SocketClient::set_query() get reset correctly if an exception happens (which used to prevent SocketClient::get_mset() ever getting called). Thu Mar 15 17:06:51 GMT 2001 Olly Betts * Oops, checked in too much - backing out the wrong stuff. Thu Mar 15 16:30:14 GMT 2001 Olly Betts * Include autoconf/definedir.m4 in tarball. Thu Mar 15 16:06:02 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Updated network protocol number to reflect change adding OP_ELITE_SET. Thu Mar 15 15:31:35 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * QuartzPositionList should no longer complain if positional information isn't available for a particular term / document combination. * Updated todo. Thu Mar 15 13:55:53 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Removed OmDocument::set_wdf() and OmDocument::add_term(). Replaced with OmDocument::add_term_nopos(), which adds a term without specifying positional information, taking an optional wdfinc parameter. Updated testsuite correspondingly. Thu Mar 15 13:14:10 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Update TODO Wed Mar 14 17:51:40 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Added new operator OP_ELITE_SET, which replaces match_max_or_terms option. Can now specify exactly which terms to apply max_or_terms to. max_or_terms is obsolete: an error will be thrown if an attempt is made to use it. Wed Mar 14 15:56:20 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Added test for and implementation of multiple XOR queries. * Fix small bugs with get_termfreq_est() for xor and andnot postlists. Mon Mar 5 16:08:01 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Catch exceptions within the loop in SocketServer - exceptions should normally not kill the server. This means that that a client still works after eg an OmDocNotFoundError, which was causing problems before. Fri Mar 2 18:15:47 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Add a new query operator, OP_WEIGHT_CUTOFF, which returns only those documents from a query which have a weight greater than a specified cutoff value. Also added OP_PERCENT_CUTOFF, but this is not yet implemented. * Added test of new cutoff function, cutoff2. * Updated network protocol to pass new queries. Increased protocol version number (to 10). * Added set_cutoff() function to OmQuery, to set cutoff parameter. * Updated TODO. * Remove unused member "PostList * postlist" of LocalSubMatch. Thu Mar 1 12:04:53 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Updates to HACKING. Wed Feb 28 18:42:47 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Fixes to neaten up configure output, and to test for the Perl Text::Format module, thanks to James Aylett (patch slightly modified, so if it doesn't work it's probably my fault). * configure.in checks for presence of ftime() and gettimeofday() * Added OmTime, a class to deal with times to an accuracy of microseconds, or whatever the best accuracy the system supports is. * Use OmTime in network code instead of time(NULL). Wed Feb 28 14:23:11 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Merge common code from OmSocketLineBuf::wait_for_data() and OmSocketLineBuf::do_readline() into OmSocketLineBuf::attempt_to_read(), and thus fix bug in wait_for_data() when 0 bytes returned from read(). Tue Feb 27 18:40:33 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Fixes to remote matcher code to cope better with failure of submatchers: - Timeout for all submatches now correctly start at same time. * Added more extensive testing of above code, involving adding new parameters to inmemory to cause abort()s at various stages of the match. * Remove obsolete (and unused) method SocketServer::read_global_stats() * Errors are now generated from strings by using omerrortypes.h with suitable #defines before including it. Updated todo accordingly. * Extend scope of try{} clause in socketserver, so that SocketServerFinished exception never escapes. * Default timeout in progserver increased to 30000. * Default timeout in tcpserver increased to 15000. Mon Feb 26 23:51:23 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Split query generation into stages: make new query, add subqueries, then finalise. Add private methods to OmQuery reflecting this, and convert constructors into template methods, allowing any iterator to be used for constructing queries. Removed reference to <vector> from omquery.h * Add omqueryinternal.cc, containing implementation of query internals. (surprise!) * OmQuery::Internal now has some helper functions to determine the properties of each particular query. These should be converted into a lookup table, and the remaining hard-coded properties (for AND_MAYBE, AND_NOT and FILTER) factored out. * window property removed from constructor. This can now be set by a separate call, to "OmQuery::set_window()" * Added a deleter_vector class, to store subquery pointers in a more exception-safe way. * Represent undefined queries by the new OP_UNDEF operation, rather than a dedicated flag - cleaned up some logic. * Some modifications to XOR handling: should now behave like OR and AND - doesn't need to be binary. (*untested*) * Fixes to serialisation code in socketcommon.cc to work with new OmQuery::Internals - this should really be done in OmQuery::Internal. * One tiny fix in tcpclient.cc - an &fdset was missed for exception conditions in a select() call. * Testsuite modified and expanded to check changes to OmQuery. Mon Feb 26 18:20:16 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Fixed some compiler warnings reported by James Aylett. Mon Feb 26 14:27:09 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Avoid waiting for a timeout when closing a SocketClient. Fri Feb 23 14:38:09 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Updated todo.xml Thu Feb 22 18:14:41 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * OmRegexFilterNode has an output for non-matching strings as well as matching. "out" output renamed to "matching". Thu Feb 22 16:56:33 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Fixed a bug in the fetch() bits: requesting documents at the same time from different OmMSet instances could cause a document to disappear from the cache at the wrong moment. Thu Feb 22 15:39:49 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Updated todo. Thu Feb 22 12:14:12 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Some small changes in quartz to improve exception safety and error checking. * Remove unwanted OmExpandWeight::get_expand_k() method. Wed Feb 21 19:02:44 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Overhauled the OmMSet::fetch() etc. methods. Documents are requested when fetch() is called, but are only fetched (all at once) when the first get_document() is called. Wed Feb 21 19:30:20 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Add an errorhandler member to MultiMatch, and remove it from various calls which used to pass it about. * Add an EmptyMatch object, which always returns an emptypostlist. * Implement errorhandlers at start of query for initially down nodes. Improve test for errorhandlers to cover many more cases. * Add many debug messages to error handling code. Wed Feb 21 14:33:50 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Use a deque<> directly instead of a queue, due to conflict with Solaris headers. * Split the collect_doc loop into a separate function * SocketClient should be more robust against request_doc() being used without collect_doc(). Tue Feb 20 18:01:13 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * SocketClient implements a queue/cache system which means that collect_doc() doesn't need to be called in the same order as request_doc(). Mon Feb 19 15:43:49 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Converted btree code to use new/delete instead of [mc]alloc/free. Mon Feb 19 13:52:23 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Added new test collapsekey2 designed to work with muscat36 databases, which have different key handling. Added two new categories of tests to cope with this. * Adjusted BackendManager's key generation, and updated test_specialterms1() accordingly. Fri Feb 16 17:26:59 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Implemented a test for keep-alives * Added "-tNNNN" option to omprogsrv to pass a timeout value in. * Added get_network_database() in apitest to set the timeout. Fri Feb 16 16:13:47 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Incremented the remote protocol version, since a new message type has been added. Fri Feb 16 15:55:31 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Added an OmDatabase keep_alive() method to gently prod any remote databases. Applications can use it to avoid the remote servers timing out between queries. Not properly tested yet... Thu Feb 15 19:05:24 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * The remote server now supports a separate timeout used while idle, rather than actually servicing a request from the client. Options --idle-timeout and --active-timeout added to omtcpsrv, and --timeout sets both timeouts. * Added a documentation comment or two. Wed Feb 14 18:43:00 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * In Muscat3.6 backends, don't read the key from the keyfile unless keyno is 0. Tue Feb 13 13:57:34 GMT 2001 Andy MacFarlane <andym@omsee.com> * Added parameter expand_k in omsettings in order to change value of weighting in expand process. Tue Feb 13 13:25:45 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Network timeouts now apply to write() operations as well as read(). * SocketLineBuf::do_writeline() now correctly deals with EAGAIN. * SocketServer has new method writeline() which handles write access to the OmLineBuf. Mon Feb 12 12:35:41 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Make OmPositionListIterators and any other appropriate iterators (postlist, termlist) have a default constructor allowing them to be declared uninitialised. Mon Feb 12 10:51:41 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Updated todo.xml Thu Feb 8 17:59:44 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Moved omnodepad.h from include/om to indexer/indexgraph/ * Updated todo.xml Thu Feb 8 17:35:45 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Updated todo.xml Thu Feb 8 17:00:02 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Quartz now has a metafile which for now has a magic string and a version number. * sys_open_to_read() and delete_file() exported from btree.cc. The latter was renamed to sys_unlink_if_exists(). * Include btree_types.h from btree_util.h Wed Feb 7 17:33:00 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Fix to InMemory databases: terms can be added even if they don't have any positions. (Was causing adddoc2 to fail). Tue Feb 6 18:06:13 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Rename OmMSet::fetch_items() to fetch(). Add an overloaded fetch() method taking a single iterator, to fetch a single item. * Fix some documentation comments. Mon Feb 5 19:16:45 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Update TODO Mon Feb 5 19:12:39 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Test behaviour of OmDocument methods add_term(), set_wdf(), remove_posting() and remove_term(). Fails for inmemory databases. Mon Feb 5 17:46:16 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Modified the behaviour of timeouts during a match. Internally, an absolute timeout time is set from start_match(). Now remote nodes should timeout at NOW + remote_timeout, rather than waiting remote_timeout milliseconds for each read(). Needs some more testing. Mon Feb 5 17:21:52 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Fix a missing #include in socketcommon.h Mon Feb 5 17:15:46 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Sort generated todo lists. Fri Feb 2 17:44:54 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Add an OmDocument::add_term() method. * Remove an unwanted check that wdf's are greater than 0: wdf of 0 _is_ allowed. Fri Feb 2 15:04:08 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Hopefully, a fix for segfaults/assertions when network errors happen halfway through a match, when handled by OmErrorHandler objects: give the place-holder EmptyTermLists weighting objects. Thu Feb 1 19:03:10 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Fixes to OmDocument interface: add parameters specifying the changes to be made to the wdf to add_posting and remove_posting, properly implement remove_posting and remove_term, including throwing of exceptions, and add a set_wdf() method to set the wdf to an absolute value. Corresponding modifications to OmDocumentTerm. * Fix commented out errorhandler code in multimatch.cc, so it still compiles. * Add a check to avoid division by zero in bm25weight.cc Thu Feb 1 15:59:30 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Add a missed #include to omqueryinternal.h Thu Feb 1 14:44:04 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Add some error handler code, with a "#if 0"ed out comment, to multimatch.cc Wed Jan 31 15:45:16 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Remote servers should close down more gracefully when the client closes down in the middle of a request. Tue Jan 30 16:52:05 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Removed unneeded #includes of STL headers from API headers. * All #includes of files which are not installed use "" rather than <>. This means all #includes of OM files except for those which are in example programs (and hence might be compiled against the installed library). "" search paths are a superset of <> search paths: hence this should ensure that the om headers work however they are used. * Moved supplied OmExpandDecider subclasses into include/om/omexpanddecider.h * Remove om_termname_list from API: not used except by code which displays it, and in testsuite. Tue Jan 30 12:45:22 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Removed #include <vector> from omstem.h Tue Jan 30 11:09:29 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * OmStem::get_available_languages() now returns a string rather than a vector. Languages are space-separated. Updated apitest. * Added definition of RTLD_DEFAULT to internaltest.cc Mon Jan 29 16:11:15 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * OmIndexerMessage's copy-on-write slightly more careful. * Improved debugging in OmIndexerNode * Fixed a bug in the termlistadd node - it tried to append to a non-vector. Fri Jan 26 16:21:09 GMT 2001 Olly Betts * indexerxml.cc: Removed some dead code. Fri Jan 26 16:11:13 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Added omindexerdescinternal.h to the Makefile Fri Jan 26 15:50:32 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Move OmQuery definition into a separate header file. Fri Jan 26 15:46:42 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Move OmQuery::OP_LEAF out of public visibility, by taking it out of the enum, making it a static const member of OmQuery::Internal, and introducing an OmQuery::Internal::op_t to store query operators in internally. We lose the ability for the compiler to warn us if we forget to check for a particular operator in a switch statement, but this can't really be avoided. Fri Jan 26 13:49:33 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Added get_output_{node,pad} to OmIndexerDesc * Implemented and testes OmNodeInstanceIterator * Implemented OmIndexerBuilder::build_from_desc, which somehow hadn't been. indextest now uses it. Thu Jan 25 18:47:08 GMT 2001 Olly Betts * Fixed various .cvsignore files (omus -> omsee, and bindings/java/.cvsignore which seemed to have been copied from the top level one at some point). * Fixed bug in OrPostList when used in a pure boolean context (uninitialised variables). * tests/internaltest.cc: removed unused reference to getopt.h. * tests/stemtest.pl.in: removed lots of bogus semicolons after closing braces. Thu Jan 25 17:53:08 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Turn debugging messages (OM_DEBUG_TYPES) off by default. * Add todo items for release to pages generated for website. Thu Jan 25 15:11:14 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Add a couple of documentation comments: we need many more of these. No method, and certainly no class, in the API should be without an explanatory comment (not least so that they're accessible from the doxygen browser). Thu Jan 25 14:51:55 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Hide the internals of OmIndexerDesc, and adding member functions to do all the necessary operations (not all yet implemented). * Indexer internals mostly deal with OmIndexerDesc::Internal * Removed '#include <vector>' from omindexerdesc.h Thu Jan 25 13:45:49 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * When setting up an RSet, allow documents to be referenced by OmMSetIterator, for convenience. * Allow OmMSetIterators and OmESetIterators to be created without initialising, also for convenience. * Include CFLAGS for STLPORT in the flags produced by omsee-config. * Change several more missed "omus"s to omsee. Thu Jan 25 13:50:06 GMT 2001 Chris Emerson <chris.emerson@omsee.com> * Changed a mention of libomus in netprogs/Makefile.am to libomsee Thu Jan 25 12:33:46 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Build only the API docs when making dist / install. Full source docs only get built if explicitly asked for. Thu Jan 25 11:36:39 GMT 2001 Richard Boulton <richard.boulton@omsee.com> * Project name changed to Omsee. Ramifications from a technical viewpoint are: - library name changed from libomus.* to libomsee.* - libomus-config renamed to omsee-config - Autoconf macro file libomus.m4 renamed to omsee.m4, macro name OM_PATH_LIBOMUS changed to OM_PATH_OMSEE and macro now defines OMSEE_CFLAGS and OMSEE_LIBS rather than LIBOMUS_CFLAGS and LIBOMUS_LIBS. configure arguments changed name, also. Wed Jan 24 15:57:10 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Improve test of OmMSetIterators equality comparisons. * Update TODO Wed Jan 24 15:11:11 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Renamed OmNodeConnection to OmNodePad. The old name was confusing. * Added OmPadIterator, and OmNodeDescriptor::{in,out}puts_{begin,end}. Also implemented OmNodeDescriptor::get_type() * Added output function for OmPadIterator * Added test for OmPadIterators Tue Jan 23 19:01:29 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Remove OmBatchEnquire from system: it may return at a later date, but for now it is simply out of date and a maintenance liability, and gives no significant advantage. Code now exists in ombatchenquire.h in CVS, but is not part of build process. * Make OmEnquire and OmMSet have reference counted internals: this introduces a further layer of indirection, but tidies up things greatly. * Remove get_docs() and get_doc() methods from OmEnquire. get_doc(docid) is catered for by OmDatabase::get_document(docid). Added OmMSetIterator::get_document() to replace get_doc on a MSetIterator. Added OmMSet::fetch_items() to replace get_docs(): it prefetches the documents (or just those specified by some MSetIterators into a cache in the Mset). * OmMSet internals now have a reference to the creating OmEnquire object: this is set up after the mset returns from the matcher. If the reference is null, the Mset was created standalone, and get_doc methods will fail. * OmMSetIterators now have a reference to the MSet, instead of a percent_factor stored. This allows the percent_factor to be lazily calculated, and also allows get_doc to work. * Add get_rank() method to OmMSetIterators. * Improve introspection on OmDocument, and fix a bug with assignment of OmDocument (wasn't assigning most of the internals). * Add a missing std:: on a string in omstopwordnode. * Fixed several documentation comments. * Add some more debugging to quartz. * Update TODO. * Add test_fetchdocs1 to apitest. Disable test_batchquery1. * Remove sign comparison error in VectorTermList. Tue Jan 23 16:24:04 GMT 2001 Olly Betts * No longer compile with -ansi under gcc - it causes problems by preventing Linux system headers defining various functions we need. * testsuite/testsuite.cc: when reporting exceptions, display exception's errno value if set. Tue Jan 23 15:13:58 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Killed OmIndexerBuilder::NodeType. Now use OmNodeDescriptor in its place, getting rid of another vector usage in the API. * OmNodeDescriptor adjusted for its new role: internals are refcounted, and some query as well as setting functions. * Cast the fourth argument to getsockopt() to void *. On Solaris that argument is char *. Tue Jan 23 11:24:13 GMT 2001 Olly Betts * Tiny tweak to da_document.cc. Fri Jan 19 18:36:15 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Removed get/set vector methods from OmSettings * Remote/prog arguments now passed in a string separated be spaces, not a vector. * split_words() moved from netutils.h to utils.{h,cc} * Some unnecessary #include <vector> lines removed from API headers * Removed get_config_vector() from OmIndexerNode * OmConstantNode no longer handles vector values, and OmStopWordNode and OmSelectItemsNode use space-separated values instead. Thu Jan 18 17:31:55 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Removed mentions of std::vector from OmIndexerMessage, and updated the relevant bits. * Fixed a bug with the copy-on-write parts of OmIndexerMessage. Thu Jan 18 17:09:43 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Fix to btreetest.cc: shouldn't fail if test data files are not available. * Fix to multimatch.cc: clamp hits values to number of documents matched after collapsing and match functors, rather than before. Thu Jan 18 14:13:21 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Fix error messages thrown by quartz when a database doesn't exist / already exists and isn't to be overwritten. * Make quartz clean up an old database directory, if database_allow_overwrite is specified, so that old base files don't confuse things. Wed Jan 17 18:18:50 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Added all used toplevel directories to DIST_SUBDIRS Wed Jan 17 18:13:50 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Martin committed changes to btree.cc which should fix some random problems experienced with quartz. * Added in a test framework for the btree code, which exhibited the bug fixed by martin until updating to his fixed version. Wed Jan 17 16:10:02 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Moved the socklen_t definition so that tcpclient.cc could see it. Tue Jan 16 20:35:18 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Fix creation of quartz databases in backendmanager: all tests should now pass. Tue Jan 16 20:28:46 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Fix quartztest so that the tables / databases get created. Implement new test "create1" to check that creation of databases works as specified. * Fix test for whether we are allowed to overwrite a database. * Implement much better logging from quartz when making new databases. * Clean up an error message from btree.cc Tue Jan 16 19:24:49 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Don't display messages about allocation failures in the first iteration: if the failures go away when repeating, we're not interested. Tue Jan 16 18:28:03 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Add two new OmSettings, for use when opening databases: database_create and database_allow_overwrite. The idea is that databases will only be created if database_create is specified, and that when they are being created an exception will be thrown if there's an existing database unless database_allow_overwrite is specified. * Implemented database_create and database_allow_overwrite for Quartz, which is the only relevant type for now. Note that this breaks the testsuite, quite badly. * Add OmDatabaseCreateError, which is thrown when creating a database fails. Tue Jan 16 17:08:47 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Fixed a typo in -ldl detection Tue Jan 16 15:15:30 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Remove OmNeedRecoveryError, and quartz's quartz_perform_recovery parameter. Quartz now always performs recovery if needed: there is no need for this added complication. * Update TODO Tue Jan 16 14:25:03 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Changed references to SIGCLD to SIGCHLD as it's more portable. Tue Jan 16 14:02:34 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Added check for whether -ldl is necessary. Tue Jan 16 11:51:25 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Added compatibility bits for early libxml1 versions. Mon Jan 15 18:18:01 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Removed use of hstrerror, which isn't portable. Mon Jan 15 18:02:42 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Fix assertion which was the wrong way round in multimatch.cc Mon Jan 15 17:46:13 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Fix initialisation of Btree_base::sequential when creating database. Mon Jan 15 15:43:46 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Renamed some uses of XML node structure for libxml2 compatibility. Mon Jan 15 15:35:50 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Add some fixes to sequential access to btrees from Martin. Mon Jan 15 12:23:02 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Set matches_* to the known value of the number of hits if we have been unable to return as many hits as were requested, and therefore know the exact number. * Update test_matches1() to check this works correctly. Passes. Fri Jan 12 17:45:27 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Implement get_doclength() for QuartzDatabase * Remove unused parameter from QuartzRecordManager::add_record() Fri Jan 12 17:14:16 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Fixed MultiAllTermsList, which had stopped working. Fri Jan 12 15:32:34 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Add test to apitest (postlist6) that doclengths got from postlists are the same as those from databases. Fails for Quartz for the moment. * Add test to quartztest (disktable3) to try and bring out a bug in adding to tables. Fri Jan 12 15:20:09 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Fix STLport macro so that it will accept an STLport installation with foo/include/stlport and foo/lib instead of foo/stlport and foo/lib. Fri Jan 12 14:04:37 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Fixed buglet in configure.in which produced a bogus warning about regex libraries. * Added an AC_MSG_RESULT corresponding to an AC_MSG_CHECKING for the libxml flags Fri Jan 12 13:18:07 GMT 2001 Chris Emerson <chris.emerson@open.muscat.com> * Fixed skip_to() in QuartzAllTermsList Fri Jan 12 12:52:57 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Turn debugging off in btree.cc: accidentally committed a version with lots of verbose debugging on. Fri Jan 12 11:16:17 GMT 2001 Olly Betts * glibc 2.2 needs `#define _GNU_SOURCE' to give us RTLD_NEXT. * C++ comments changed to C comments in malloccheck.c. Thu Jan 11 16:42:07 GMT 2001 Olly Betts * Removed unnecessary `#include "alltermslist.h"' from d[ab]_database.h. Wed Jan 10 14:28:10 GMT 2001 Olly Betts * Updated various .cvsignore files. Tue Jan 09 19:03:12 GMT 2001 Olly Betts * Tidied up dead code from OmAllTermIterator. Tue Jan 09 18:41:21 GMT 2001 Olly Betts * OmAllTermsIterator merged into OmTermIterator. * When iterating over all terms, skip_to() doesn't set at_end() when it skips off the end. Fixed except for quartz and added regression test (allterms3). Tue Jan 9 18:14:04 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Add some extra debugging code into btree code in quartz. Tue Jan 9 17:22:12 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Make skip_to() work (ie, do nothing) on OmTermIterators, when the term iterator is at_end() already. Tue Jan 9 13:32:08 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Updates to make the quartzdump utility more useful. * Added new test to test cursor behaviour in quartz. Tue Jan 9 11:31:26 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Remove an item from TODO Tue Jan 9 10:27:35 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Add a missing \ in tests/Makefile.am Mon Jan 8 19:03:42 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Increment format number in quartz/btree_base.cc. This isn't actually due to a change in the btree format, but to ensure that some databases which were causing problems are rebuilt. We need to add quartz format numbers, separate from the btree format numbers, to check this kind of thing. Mon Jan 8 15:13:04 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Add some debugging, and fix a test data file. Mon Jan 8 13:49:50 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Fix failure to initialise OmMSet::Internal::have_percent_factor. Thu Jan 4 17:21:24 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Add new test for a postlist with many documents, designed to fail and exhibit the bug andy has found. So far doesn't fail. Wed Jan 3 19:30:30 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Add get_percent method to OmMSetIterator. Untested. * Add om_percent type to om_types. * Add test of sort order of terms in get_matching_terms. Tue Jan 2 15:05:53 GMT 2001 Richard Boulton <richard.boulton@open.muscat.com> * Move documentation of parameters for OmSettings out of omsettings.h and into docs/. Not in distribution at the moment: it is awaiting placement into a permanent home (I think the user manual is the correct place). * Add a bm25weight_min_normlen parameter: this specifies a cutoff on the minimum value that can be used for a normalised document length: smaller values will be forced up to this cutoff. This prevents very small documents getting a huge bonus weight. Wed Dec 20 16:54:41 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fix warning in QuartzTable Wed Dec 20 15:13:05 GMT 2000 Olly Betts * OmTermListIterator merged into OmTermIterator. * Added OmRSet::contains() - checks if an RSet contains a given docid. * Methods returning `const om_termname' now just return `om_termname'. Tue Dec 19 16:55:02 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Implemented AllTermsIterator for DA and Multi databases. Added a test which actually uses a multidatabase, but disabled it as it currently fails for both Quartz and DA. Tue Dec 19 12:14:05 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Adjusted TODO Tue Dec 19 20:06:43 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Escaping for termnames in remote database communication modified to use only characters in range 33 to 126. Should now work on all architectures. * Increased OM_SOCKET_PROTOCOL_VERSION to 8. Mon Dec 18 17:23:48 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Added "match_cutoff" option to get_mset OmSettings. * Added test for "match_cutoff" to apitest. * Updated todo.xml Mon Dec 18 17:19:40 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Implemented opening positionlists from inmemory databases. Added a testcase as well, which is run for quartz and inmemory. Mon Dec 18 15:48:16 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Implemented AllTermsIterator for quartz. Interaction with QuartzPostList needs to be cleaned up. * Added new functions [un]pack_string_preserving_sort(), which pack a string in a way which preserves sort order. Added a testcase to quartztest. * Now use [un]pack_string_preserving_order() on postlist keys, so that we can get at terms in sorted order. (And get_tname_from_key() added to separate out the term-unpacking as well as the packing) * Enabled test_allterms1 in apitest, but put it in a group of tests only run for inmemory and quartz. This should be extended to other backends. Mon Dec 18 14:33:42 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Implemented tests of msetiterator and esetiterator copying and assignment. * Fix segfault when assigning to [me]setiterator to end. * Updated todo.xml Mon Dec 18 13:52:53 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Move creation of first key in a quartz postlist into a named function (was scattered through the code). Mon Dec 18 11:19:53 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Update todo.xml Mon Dec 18 10:47:42 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Remove const from reverse_iterators in api_db.cc, so get around brokenness of egcs 1.1.2 Mon Dec 18 09:41:54 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Added apitest_space.txt to the Makefile.am Sun Dec 17 16:15:03 GMT 2000 Olly Betts * Replaced all API uses of om_termname_list with OmTermIterator. Fri Dec 15 18:54:10 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Implemented OmAllTermsIterator interface in API. It works for inmemory - the rest haven't been implemented yet. The test is commented out. Fri Dec 15 15:47:53 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Updated TODO. * database returns positionlists as AutoPtrs; this neatens some of the code from the previous changelog entry, and makes explicit the ownership of the pointer. postlists and termlists should be AutoPtrs, likewise. Fri Dec 15 14:59:41 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Make OmPositionListIterators retrieved from OmPostListIterators valid indefinitely (were only valid until OmPostListIterator was moved). This makes them consistent with OmPositionListIterators retrieved from OmDatabases. * Renamed PostList::get_position_list() to read_position_list. Added PostList::open_position_list() which opens a new positionlist and returns it as an AutoPtr. Fri Dec 15 13:09:51 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Make TODO.release file, containing TODO items for next release. Fri Dec 15 12:06:42 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * QuartzDatabases now autoflush after 1000 changes. Fri Dec 15 11:40:37 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fix segfault in quartz_table Fri Dec 15 11:23:15 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Removed begin/end_session() from quartztest Thu Dec 14 19:19:40 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Done todo item: For writable databases, make the behaviour currently obtained by calling begin_session() and later end_session() the default behaviour. If users want the changes to be applied immediately following a change, they should call flush(). Thu Dec 14 17:48:27 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fix problem with recalc_maxweight() not being called before a get_maxweight() in multimatch.cc if a node prunes at the top level, since recalculate_w_max wasn't being checked at this point. Introduced helper function getorrecalc_maxweight() to tidy this up. * Tidy up Asserts in common/netutils.h. Thu Dec 14 17:11:17 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Adjusted todo.xml Thu Dec 14 16:21:27 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Removed handling of OmDatabaseModifiedError in the API - it's up to the users to know what to do. Disabled quartzoverwrite2 as it's now inappropriate. * Added OmDatabase::reopen() to help them do so. * Adjusted todo.xml Thu Dec 14 14:38:39 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Added another quoting regression test. Thu Dec 14 14:36:53 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Fixed the quoting _again_, since it broke with backslashes. Thu Dec 14 12:49:58 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fix problems with recalculation of maxweight in postlists, causing AssertParanoid at multimatch.cc:393. When a prune was happening, recalc_maxweight flag was not always being set, because some {next,skip_to}_handling_prune() methods weren't being passed the matcher, and a default argument of 0 was being used. I've now removed the default argument and pass the matcher always. Sometimes the recalc could perhaps be avoided, but this is certainly safer for now. Thu Dec 14 12:30:02 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Sleepcat is now disabled by default. * Fixed a problem with tabs not being quoted in the remote case. All control characters are now quoted. Incremented the protocol ver number. Thu Dec 14 11:54:17 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Added regression test to check that all characters can be are safe to exist in document keys. * Fixed OmDebug so that it can display messages containing zero bytes (use fwrite instead for fprintf) Wed Dec 13 18:43:40 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Add regression tests to apitest to check that OmMSet and OmESet begin() and end() iterators compare equal if the mset is empty. Fix matchfunctor used in test_matchfunctor1 so that it works for Muscat3.6 databases. Wed Dec 13 18:35:05 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Fixed bugs in OmMSet and OmESet iterators: begin() when the set was empty produced an invalid iterator, and the assignment operators would have failed when end iterators were involved. * Added a check in test_matchfunctor1() that the mset returned is not empty. Wed Dec 13 18:18:09 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Turn off tests involving terms with newlines or zero bytes in terms with Muscat3.6 backends; these can't be expected to work. Wed Dec 13 18:06:29 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Uncommented the code in pctcutoff1 - test now passes. Wed Dec 13 17:53:19 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fix: m36_key_file parameter wasn't being set with DA flimsy databases (thanks to Mark Hagger). * indextest.cc changed to use "\n" instead of endl - compiles with STLport again. Wed Dec 13 17:27:17 GMT 2000 Olly Betts * OmMSet is now an STL compatible container. * Om*Iterator::difference_type was unsigned - now signed. * net/readquery.ll: fixed compiler warning. * Remote backend was defaulting to collapse on key 0. Wed Dec 13 16:48:49 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Add handling of DatabaseModified to OmEnquire::get_mset() Wed Dec 13 16:28:47 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Escape zero bytes in the network protocol, since these were breaking things when being passed across as part of keys. * Add test of having zero bytes in the collapse keys. * In testsuite, only display the context of errors if there is one present. Wed Dec 13 14:44:46 GMT 2000 Olly Betts * Implemented OmMSet::operator[](). Wed Dec 13 14:30:57 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Added support for handling overwritten database conditions. Currently works for OmDocument::get_{data,key}(). Wed Dec 13 13:20:45 GMT 2000 Olly Betts * Elaborated API hitlist in todo.xml. Wed Dec 13 12:28:18 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Fix to BackendManager to stop it removing tables prematurely. Wed Dec 13 11:49:53 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Apply the stemmer to query terms used in spaceterms1. (Not doing so caused the test to fail spuriously on local databases, since the terms are stemmed at index time). The test still fails with da and db databases. * Fixed a problem when stats were unpacked from the network stream, where term names were not correctly unquoted. This led to double quoting of global statistics. Wed Dec 13 02:49:05 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Implement copy and assignment methods for OmPositionListIterator. Wed Dec 13 02:37:14 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * ESet and MSet iterators now cope with being copied / assigned when the internals are null. * Added copy method to OmPositionListIterator interface. No implementation for this yet though, or for the already existing assignment method. * Added a couple of easy items to the todo list to check regarding iterators. Wed Dec 13 02:19:50 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fix small bug in mset_range_is_same{,_weights}() Wed Dec 13 02:07:59 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * And add copy and assignment for OmESet, similarly. Wed Dec 13 01:58:38 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Add copy and assignment operators to OmMSet so that the internals get copied (rather than just the pointer to them). Really want a small internals class, holding a refcount to the real internals, or else to use refcntptrs for the members of OmMSet::Internal which are potentially large (eg, items) Wed Dec 13 01:05:44 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Add Asserts to api/omdatabaseinternal.cc to ensure that there is at least one database present when opening termlists, etc. This should really be done by throwing errors. * Add destructor to OmMSet, deleting internals. * Implement getting positionlists from QuartzDatabases. * Fix bug in copy constructor of OmDocument::Internal - some fields were not being copied. * Fix api_db.cc so it compiles: code referring to mset by subscripting temporarily commented out. Tue Dec 12 20:05:10 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fix bug with truncation of intermediate keys in btrees. Tue Dec 12 18:33:12 GMT 2000 Olly Betts * OmMSet::get_docs() now takes sane parameters. * Various test programs updated to nearly work with new OmMSet interface. Tue Dec 12 17:18:33 GMT 2000 Olly Betts * Finished cleaning up OmMSet; updated TODO to reflect this. Tue Dec 12 17:16:02 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fix bug with sorting keys in btrees. * Various fixes to quartzdump. Tue Dec 12 15:40:28 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Nettest compiles with new OmMSet interface. Tue Dec 12 15:09:14 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Fixed an endless loop in QuartzWritableDatabase::do_delete_document. Tue Dec 12 15:07:39 GMT 2000 Olly Betts * OmMSetItem replaced by OmMSetIterator. * Added OmESet::empty(). * Fixed == on OmESetIterator. Tue Dec 12 14:16:57 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Renamed arguments to compare_keys() Tue Dec 12 12:05:24 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Added quartzdump utility to dump the contents of a quartz table. Tue Dec 12 10:54:44 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Added constructors and destructor to OmRSet. Mon Dec 11 16:32:07 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Remove caching of average length in OmDatabaseInternal, was breaking asking for the average length of a database which was being modified. Mon Dec 11 16:03:46 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * If no documents in database, don't divide by zero in omdatabaseinternal.cc * If a document object doesn't come from the database, when iterating through its termlist don't open postlists from the nonexistent database; open them from the DocumentTerm objects in the document instead. * Remove old unused and undefined method OmDocument::Internal::add_posting() Mon Dec 11 13:53:18 GMT 2000 Olly Betts * omenquire.h: Removed superfluous declaration of OmMSetCmp. Mon Dec 11 13:38:12 GMT 2000 Olly Betts * Updated TODO list. Mon Dec 11 13:24:45 GMT 2000 Olly Betts * OmRSet internals encapsulated. Mon Dec 11 12:42:02 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * OmIndexerMessage::get_element and ...::operator[] now have non-const versions. Fixes const-warning in omstemmernode.cc. Mon Dec 11 12:17:32 GMT 2000 Olly Betts * Better typedef-s for Om*Iterator::difference_type. Sun Dec 10 17:51:41 GMT 2000 Olly Betts * Corrected various typedef-s for Om*Iterator. Sun Dec 10 17:28:59 GMT 2000 Olly Betts * Updated TODO to reflect OmESet being cleaned up. * Term position argument to OmDocument::add_posting() is now optional (as it was for OmDocumentContents). Sun Dec 10 16:55:17 GMT 2000 Olly Betts * OmESetItem is dead - long live OmESetIterator! * net/socketclient.cc: fixed compiler warning. * net/socketserver.cc: fixed compiler warning. Sun Dec 10 13:29:15 GMT 2000 Olly Betts * Started to encapsulate OmESet/OmESetItem interface. Interface is close, but currently implementation is exposed in omenquire.h. * apitest wasn't running puncterms1 or spaceterms1 - fixed. Sun Dec 10 11:34:39 GMT 2000 Olly Betts * om/autoptr.h -> autoptr.h in lots of places. Fri Dec 8 17:20:04 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Updated indextest to not use AutoPtr<OmIndexer> Fri Dec 8 17:14:09 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Oops - removed one too many #include "autoptr.h" Fri Dec 08 16:50:42 GMT 2000 Olly Betts * Updated TODO. Fri Dec 8 16:44:26 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * IndexerBuilder now returns OmIndexer objects instead of AutoPtr<OmIndexer>. OmIndexer's internals are reference counted. * Moved autoptr.h from include/om to common/, since it's no longer used in the API. Fri Dec 8 16:22:48 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Fixed omstemmer node bug and made debug slightly more verbose in indextest.cc. Fri Dec 8 16:13:36 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Added a test for OmIndexerMessage to indextest * Added output operator for OmIndexerMessage * Fixed a bug introduced into omvectorsplit node * Changed couts in indextest.cc to touts. Fri Dec 08 14:53:24 GMT 2000 Olly Betts * quartztest: removed OmDocumentTerm reference so that it'll compile (tests fail though). Fri Dec 8 14:42:46 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * OmIndexerMessage is now a class by itself, replacing OmIndexerData and the typedef to AutoPtr<OmindexerData>. It's now a copy-on-write object. * The indexer stuff all changed. Fri Dec 08 14:24:02 GMT 2000 Olly Betts * omindexdoc.{cc,h}: no longer used so removed. * Pushed OmDocumentTerm from API level into the internals of the library. * Updated TODO. Fri Dec 08 12:08:49 GMT 2000 Olly Betts * delve: if just given a database (or list of databases), report the number of documents and average document length. Thu Dec 7 20:39:24 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fix bug: Om*Iterators which were created over empty lists didn't have null internals, causing them not to compare equal to end iterators. * Added many asserts to iterator code. Thu Dec 7 20:02:43 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Removed OmTermListIteratorMap - OmTermListIterator can now be across either a map or an actual termlist. * Om*Iterator classes all now delete internals when reaching end, freeing resources sooner and making the comparison operators much simpler. Thu Dec 7 18:33:03 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Updates to TODO Thu Dec 7 18:32:39 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Implementation of write locks for Quartz Thu Dec 7 18:28:55 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Fix to backendmanager which doesn't return writable databases unnecessarily (which breaks things when there are write locks) Thu Dec 7 18:25:50 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Added test for Quartz write locks Thu Dec 7 17:56:19 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Add OmDocument::set_data() method which takes a string, instead of an OmData object. This allows, implicitly, use of const char *'s as well. * Further fixes to quartztest. It now compiles (though I've not tried linking it yet...) :) Thu Dec 7 17:49:50 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Fixed omkeylistiterator return types. Thu Dec 7 17:27:57 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Moved get_document() from OmWritableDatabase to OmDatabase. * Added -> operator to OmKeyListIterator. Fixed return values and constness of Om*Iterator::operator *() methods. * Some fixes to quartztest.cc, but not enough to get it compiling. Thu Dec 07 16:21:28 GMT 2000 Olly Betts * Some work on getting quartztest working. Thu Dec 07 16:10:00 GMT 2000 Olly Betts * The library now builds, but "make check" still fails. Thu Dec 7 15:39:19 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fixes to testsuite/index_utils.cc to not segfault. Thu Dec 7 14:31:56 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Fixes to api_db.cc so that it compiles. Wed Dec 6 18:50:48 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Add test for terms with spaces, newlines, zero bytes, or backslashes. * Updates to TODO Wed Dec 06 18:18:16 GMT 2000 Olly Betts * Woohoo! Now builds with just --disable-indexer! (still doesn't pass any tests). Wed Dec 06 17:24:19 GMT 2000 Olly Betts * Minor fix. Now builds with --disable-indexer --disable-quartz-backend (but doesn't pass any tests). Wed Dec 06 16:35:34 GMT 2000 Olly Betts * Fixed up lots of build errors. Wed Dec 06 14:56:59 GMT 2000 Olly Betts * Begin the long job of replacing OmDocumentContents with OmDocument. Won't currently build - expect the tree to be broken for a few days. As usual, if you want a working build, use the "snapshot_latest" tag. Wed Dec 6 12:03:44 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Backed out yesterday's overwritten-handling stuff. Wed Dec 6 11:36:55 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Increase sleep in tcpclient1 to 3 (did tcpmatch1 last time, duh!) Tue Dec 5 17:13:38 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * QuartzDocument now has a QuartzTableManager pointer rather than two QuartzTable pointers. * QuartzDatabase::do_get_document_internal() copes with Btree overwritten conditions. * QuartzDocument methods cope with Btree overwritten conditions * New method QuartzTableManager::reopen_tables_because_overwritten() for helping with the above. * New test in quartztest for testing overwritten problems from the API rather than just internally to Quartz. * Fixed a typo in omsettings.h comment (quartz_dir used to refer to a sleepycat database) Tue Dec 5 16:53:46 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Update todo. * Fix undesirable assert in stats.h Tue Dec 5 15:41:40 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Replace OmMSet::docs_considered with OmMSet::matches_{lower_bound,estimated,upper_bound}, giving more useful information. * PostLists now have get_termfreq_{min,est,max}(), instead of get_termfreq(), so the postlist tree can be used to calculate the matches_* values for the mset. Implemented for all the postlist subclasses. LeafPostList's still has a virtual get_termfreq() method, and implements get_termfreq_* in terms of it, so database postlists don't need to be modified. * Passing of MSets across network modified to pass the new information. * PendingMSet uses the values from the mset to return the values of get_termfreq_*(). This assumes that recalc_maxweight has been called first, which is has been. * Muscat36 databases get_termfreq() methods fixed to use a LeafPostList rather than a PostList. * Renamed test docs_considered1 to matches1, and add several more cases to it. Tue Dec 05 15:38:27 GMT 2000 Olly Betts * Added tests/testdata/apitest_punc.txt to distribution. Tue Dec 05 15:07:26 GMT 2000 Olly Betts * Oops, failed to add new file `tests/testdata/apitest_punc.txt' with last check-in. Tue Dec 05 14:34:04 GMT 2000 Olly Betts * OmTermListIterator implemented for an OmDocument with terms stored in a map rather than coming from a Document. * Fixes to remote protocol so solve problems with terms with punctuation in (regression test punc_terms1). Mon Dec 4 15:02:53 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Makefile tweak in docs/ Mon Dec 04 14:13:05 GMT 2000 Olly Betts * More OmDocument work. Mon Dec 4 11:43:46 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Renamed OmMSet::mbound to docs_considered. Mon Dec 4 10:11:11 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Added some more std:: prefixes for STLport. Fri Dec 1 17:45:01 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Increase the sleep after the server is started by tcpclient1 to 3 seconds so that the server is more likely to have started when then test ran. This seems to be what was causing the test to randomly fail. Really, we want a better solution to this (such as the server picking a port and writing it out on stdout once it is listening.) Fri Dec 1 17:27:02 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Swapped the ports used by tcpclient1 and tcpmatch1 to see if the random failures follow suit. Fri Dec 1 17:21:07 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Another minor fix to work with STLport. Fri Dec 1 17:01:02 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Add std::'s and similar to fix compilation with STLport. * Fix bug in expander due to nth_element being called on the wrong element. This is the counterpoint of the bug fixed on Nov 24th with the matcher. Fri Dec 1 16:30:26 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Minor tweaks in error strings. Fri Dec 01 12:57:59 GMT 2000 Olly Betts * Corrected erroneous documentation comment. Thu Nov 30 18:50:16 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * OmSettings no longer throws exceptions internally (at the cost of slightly more code duplication). Thu Nov 30 18:05:39 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Add a few std::'s to strings which had been missed. Thu Nov 30 16:41:11 GMT 2000 Olly Betts * OmDocument::Internal now knows its OmDatabase. Thu Nov 30 16:40:33 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Add context to network errors thrown from OmSocketLineBuf. Thu Nov 30 16:21:47 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Added a quartz test for largish bitmaps. * Fixed the code so that the above test would pass. Thu Nov 30 16:10:37 GMT 2000 Olly Betts * OmDocument::add_posting() pretty much done. Thu Nov 30 15:51:46 GMT 2000 Olly Betts * More work on OmDocument::add_posting(). * Added OmTermListIterator::positionlist_begin() and OmTermListIterator::positionlist_end(). Thu Nov 30 14:16:28 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Fixed a prototype bug introduced with the last commit. Thu Nov 30 13:23:14 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Bcursor::find_key now more const-correct. * Bcursor_create now returns an AutoPtr. The resultant changes fixed a memory leak. * New exception: OmDatabaseModifiedError. Thrown when an open revision of a table is no longer valid due to multiple updates. * quartzoverwrite1 test re-enabled, and passes. Thu Nov 30 13:23:27 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * nettest now uses quartz backend: should always run all tests now (apart from tcpdead1, which doesn't work). Wed Nov 29 19:46:22 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * In testsuite, report the context of Om exceptions as well as their type and message. Wed Nov 29 18:30:10 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Add "remote_connect_timeout" parameter to be supplied when opening a remote database, to specify a separate timeout for the connect operation to that used once the connection has been opened. * Add setting of the context of errors thrown by tcpclient or progclient, so that the node which failed can be determined. Wed Nov 29 17:50:03 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Reworked Btree_base::read() a bit so that it's hopefully more readable. * Hopefully fixed a bug when the bitmap approaches 1k. * Fix to catch problems re-opening databases after an apply in Quartz. Wed Nov 29 17:30:26 GMT 2000 Olly Betts * Added OmDocument::termlist_begin() and OmDocument::termlist_end() Wed Nov 29 16:53:13 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Comment out quartzoverwrite1 test so that it doesn't cause problems for other developers. Wed Nov 29 16:27:19 GMT 2000 Olly Betts * docid now in Document class rather than subclasses. Wed Nov 29 14:54:13 GMT 2000 Olly Betts * leafdocument.cc renamed to document.cc * Document class now has a pointer to the associated Database class so we can open termlists, etc. Wed Nov 29 14:03:43 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Add quartztest to test for case when a database block gets overwritten. Currently, this fails because an exception doesn't get thrown when it should. Wed Nov 29 13:18:24 GMT 2000 Olly Betts * LeafDocument class renamed to Document. Wed Nov 29 12:35:49 GMT 2000 Olly Betts * Added OmDocument::remove_posting(), OmDocument::remove_term(), and OmDocument::clear_terms() Wed Nov 29 12:31:22 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Added another revision number in the middle of the base file, before the bitmap. Also now check that there isn't any extra data at the end after we finish reading it. Incremented the database format. Wed Nov 29 11:55:03 GMT 2000 Olly Betts * Added OmDocument::remove_key() and OmDocument::clear_keys() Tue Nov 28 18:34:57 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Btree bitmap now merged with base - one object, one file. Tue Nov 28 17:13:56 GMT 2000 Olly Betts * Added OmDocument::set_data() Tue Nov 28 16:47:06 GMT 2000 Olly Betts * Started to alter OmDocument to be a general purpose document class which includes the functionality currently in OmDocumentContents. Tue Nov 28 15:08:38 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Added OmSplitOnCharsNode, which does a subset of OmRegexSplitNode but less slowly. * Modified RegexSplit node and regexcommon so that less copying of strings is needed. Tue Nov 28 13:42:41 GMT 2000 Chris Emerson <chris.emerson@open.muscat.com> * Moved the Btree's bitmap into a separate object. Tue Nov 28 13:06:36 GMT 2000 Olly Betts * Eliminated OmDocumentParams. * Removed unnecessary mutex from OmDocument::Internal. Mon Nov 27 18:45:27 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Implement get_position_list() on databases. Untested. Also needs some work to make consistent with open_position_list() on posting lists: see todo list. Mon Nov 27 18:29:00 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Change #define LEVEL ... to #define GET_LEVEL ... in btree_util.h: fixes problem reported by includetest Mon Nov 27 18:20:42 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Added a set of modifications to the todo.xml. Also, add a "release" field, recording the release that the work must be performed by. Mon Nov 27 18:08:35 GMT 2000 Olly Betts * A RefCntPtr now copes with being assigned to itself (used to delete the underlying object if the reference count was 1). Added regression test (refcnt2). * Now overwrite pointer inside RefCntPtr before deleting the pointed to object so there's no window where another thread can access the pointer and get the just-deleted object. Mon Nov 27 14:36:59 GMT 2000 Olly Betts * OmError class now optionally stores an errno value. Mon Nov 27 13:04:18 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Quartztest builds its test databases in a subdirectory, which is now in .cvsignore, so we shouldn't get cvs complaining about so many unknown files any more. Mon Nov 27 12:53:16 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Another TODO item. Mon Nov 27 12:15:53 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Added a couple of new items to the todo list. Fri Nov 24 19:23:56 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Quartztests now deletes all old tables, so that database formats don't cause problems. Fri Nov 24 19:03:59 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Added std:: prefix to lots of strings which didn't have it. I've probably added some where they shouldn't be: there were rather a lot of cases of this. Fri Nov 24 18:27:06 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Major bug fix! When removing elements, matcher was performing nth_element() on the first element to be thrown away, rather than the lowest element to keep. This meant that, after throwing away low scoring items, items.back() was a random item rather than the lowest scoring item. This caused the min weight for getting into the mset to go up faster than it should have done, and generally broke everything. It will be interesting to compare evaluation measures before and after this fix. Fri Nov 24 15:34:17 GMT 2000 Richard Boulton <richard.boulton@open.muscat.com> * Declare this to be version 0.4.1
Generated by dwww version 1.15 on Sun Jun 16 21:37:14 CEST 2024.